"Tory MP James Gray is dropped from role at St John Ambulance 'after saying "They all look the same to me" about Asian ministers Nadhim Zahawi and Sajid Javid' at reception
Tory MP for North Wiltshire James Gray made remark at charity bash Mr Gray, 66, is St John Ambulance commander and was at Parliament reception It was held to recognise 'extraordinary efforts' of volunteers and front-line staff He introduced Mr Zahawi to the stage as the Health Minister but was corrected Mr Gray then was said to comment 'They all look the same to me' to audience He acknowledged it was not appropriate and Mr Zahawi spoke to him privately Last night St John Ambulance said it did not tolerate racism in any form It has now asked him to stand down from all charity activities at once Mr Gray has denied saying remark and insists it was identity mix-up not racism"
A combination of complacency, lethargy and Woke crap - ‘omg GPT3 might be racist’ - means the west has handed the race to AI to China, and it may already be too late to catch up.
If China dominates AI it dominates the world like no power before it
You should have learnt by now that most people aren’t interested in the really big stories.
We’ve been directly told in the last year that there is definitely ultra tech in our skies and oceans, which either belongs to adversaries of the West or non human intelligence / life forms. And everyone shrugged.
People aren’t going to listen too hard to a senior Pentagon official if he says the US has surrendered technological dominance to China and that a point will be reached (or may already have been reached) when their lead will be insurmountable. Forever.
Cognitive dissonance innit. Much more comfortable to talk about IDS’s majority instead.
Some people are INCREDIBLY resistant to the idea of artificial general intelligence. I have an extremely smart brother who is always open to new ideas but he just won't accept that this - machine intelligence - can ever happen, let alone that it is actually happening right now
Existentially, it frightens him
Or alternatively: your extremely smart brother is actually smarter than you, and realises that everything you rave about is actually fairly sh*t smoke 'n mirrors. and nowhere near 'intelligence' (*). You are staring open-mouthed, dribbling in amazement, as someone performs the three-cup trick.
I remember an august and much-missed member of this board saying that lorry drivers would not be needed due to autonomous driving. That must have been seven or eight years ago now, and we're nowhere near. In fact, we're now suffering from a shortage of drivers. Can you recall him?
The big problem with AI is the money being swilled into the trough. Billions are being spewed at it, and they need to show results. Hence smoke 'n mirrors. Just ask Musky baby where his coast-to-cast drive in a Tesla is - promised five years ago for four years ago. Yet his current tech cannot even detect emergency vehicles...
Machine learning has many uses. But they are limited in scope, and nowhere near a general intelligence. IMO that will require a massive breakthrough in tech, not the stuff we're doing atm.
You'd be amazed at an Eliza produced by drunken first-year undergrads at the West of Scotland Uni...
(*) However you define that.
Yawwwwwwn
See below. In the next ten years we will create a computer that sails through the Turing Test, at that point everything you say here will be rendered irrelevant, whether it is true or not. That *will* be AI, because we will not be able to distinguish it from human intelligence (except that it might be much cleverer and faster)
That's why the Turing Test is such a stroke of genius. It sidesteps all the philosophy and gives you a practical threshold
I don't think that this will happen. The Turing test isn't that interesting anyway. It's when a computer claims itself to be sentient that the fun starts.
This is amazing. Arguably, this is a computer passing a *kind* of Turing Test. This is Google Duplex booking a haircut, the lady at the salon believes she is talking to a human, the appointment is made, the call ends. Turing Test "passed".
But is it really passed? This guy thinks "no, not really" and I agree with him. However, I also agree with his conclusion:
"There is no way possible that we will not have a general conversational AI in the next 10 years that can speak to any human in any language about every possible topic."
With all due respect, conversation AI is not that hard, it's just giving plausible responses to inputs.
But try explaining set theory to a computer, and then get it to give you examples back.
The difference between *learning plausible responses from a giant corpus* and *learning concepts from conversation* is as wide as the Atlantc.
Now, will we get there? Of course we will. Will it happen in the next ten years? Highly, highly unlikely. Indeed, even assuming exponential intelligence growth, it may well be fifty years away.
But that is how you pass the Turing Test. The computer can only pass the tests given to it. You're shifting the goalposts
And you're also missing the point. If a computer can persuade us it is intelligent - even human - in its interactions with us, then it is, to all intents and purposes, intelligent. That's the point of the Turing Test. Is it ACTUALLY intelligent? Conscious? Thinking? Probably not, but who knows, and Immanuel Kant would struggle to give a definitive answer. And we can no more delve into its wires to seek the answer than we can open up a human brain and locate the place of consciousness
And of course once AI reaches the stage of being indistinguishable from "true" intelligence the answer to all this is, in some senses, immaterial. There will be beings walking or talking amongst us and we will not know if they are human, or not
This is going to have momentous effects, especially online
Up until last year there was a wonderful thing called the Loebner Prize, which was basically the Turing test. And it awarded a prize for the chat bot best able to fool a human they were human.
If you read the conversations, even the ones from 2006/2007, you'll be amazed how good the responses were. Why? Because they were basically doing the same thing as GPT3 - based on the previous five words, what is the most likely sixth word?
That's not a hard problem to solve. Given billions of gigabytes of searchable text, people can easily create systems that give plausible responses. Because all they are really doing is trying to work out what the next word (or sentence) is likely to be.
But even though this is generally considered to be the best (most Turing winning) bot out there, it's still very little better than the bots from 2006/7.
Because the next stage is really hard. GPT3 is not generalised intelligence. Deep Mind is not generalised intelligence. They are really exciting, to be sure, but they don't do what you think they do.
I know what they fucking do. They autocomplete. That's it
But you seem to think that constitutes intelligence.
It doesn't.
Who the fuck knows what truly constitutes intelligence, let alone "self awareness". It is a mystery that has perplexed philosophers for thousands of years, I have a feeling Texas car park software salesman Bobby Smithson has not personally cracked the conundrum which has confounded the greatest human minds since Aristotle
"Tory MP James Gray is dropped from role at St John Ambulance 'after saying "They all look the same to me" about Asian ministers Nadhim Zahawi and Sajid Javid' at reception
Tory MP for North Wiltshire James Gray made remark at charity bash Mr Gray, 66, is St John Ambulance commander and was at Parliament reception It was held to recognise 'extraordinary efforts' of volunteers and front-line staff He introduced Mr Zahawi to the stage as the Health Minister but was corrected Mr Gray then was said to comment 'They all look the same to me' to audience He acknowledged it was not appropriate and Mr Zahawi spoke to him privately Last night St John Ambulance said it did not tolerate racism in any form It has now asked him to stand down from all charity activities at once Mr Gray has denied saying remark and insists it was identity mix-up not racism"
It is very odd however you look at it because Nadhim Zahawi looks nothing like The Saj. Even the most extreme racist would have no trouble telling them apart. It tells of impaired judgement first to get them mixed up and secondly to remark they all look the same. Gray should consider his position not because he is racist but because he is clearly not thinking straight.
"Tory MP James Gray is dropped from role at St John Ambulance 'after saying "They all look the same to me" about Asian ministers Nadhim Zahawi and Sajid Javid' at reception
Tory MP for North Wiltshire James Gray made remark at charity bash Mr Gray, 66, is St John Ambulance commander and was at Parliament reception It was held to recognise 'extraordinary efforts' of volunteers and front-line staff He introduced Mr Zahawi to the stage as the Health Minister but was corrected Mr Gray then was said to comment 'They all look the same to me' to audience He acknowledged it was not appropriate and Mr Zahawi spoke to him privately Last night St John Ambulance said it did not tolerate racism in any form It has now asked him to stand down from all charity activities at once Mr Gray has denied saying remark and insists it was identity mix-up not racism"
It's entirely in character - he created a story just a few weeks ago with a joke about a bomb being sent to another MP, and seems to crop up every few years that way. He doesn't even appear to have the pseudo notoriety of some other perennial backbenchers.
"Tory MP James Gray is dropped from role at St John Ambulance 'after saying "They all look the same to me" about Asian ministers Nadhim Zahawi and Sajid Javid' at reception
Tory MP for North Wiltshire James Gray made remark at charity bash Mr Gray, 66, is St John Ambulance commander and was at Parliament reception It was held to recognise 'extraordinary efforts' of volunteers and front-line staff He introduced Mr Zahawi to the stage as the Health Minister but was corrected Mr Gray then was said to comment 'They all look the same to me' to audience He acknowledged it was not appropriate and Mr Zahawi spoke to him privately Last night St John Ambulance said it did not tolerate racism in any form It has now asked him to stand down from all charity activities at once Mr Gray has denied saying remark and insists it was identity mix-up not racism"
"Tory MP James Gray is dropped from role at St John Ambulance 'after saying "They all look the same to me" about Asian ministers Nadhim Zahawi and Sajid Javid' at reception
Tory MP for North Wiltshire James Gray made remark at charity bash Mr Gray, 66, is St John Ambulance commander and was at Parliament reception It was held to recognise 'extraordinary efforts' of volunteers and front-line staff He introduced Mr Zahawi to the stage as the Health Minister but was corrected Mr Gray then was said to comment 'They all look the same to me' to audience He acknowledged it was not appropriate and Mr Zahawi spoke to him privately Last night St John Ambulance said it did not tolerate racism in any form It has now asked him to stand down from all charity activities at once Mr Gray has denied saying remark and insists it was identity mix-up not racism"
It is very odd however you look at it because Nadhim Zahawi looks nothing like The Saj. Even the most extreme racist would have no trouble telling them apart. It tells of impaired judgement first to get them mixed up and secondly to remark they all look the same. Gray should consider his position not because he is racist but because he is clearly not thinking straight.
Since, as you note, they don't look remotely alike, I could believe it was a genuine confusion on his part and the comment was an attempt at a joke. But that requires him to lean in hard with a 'I'm stupid' defence, since it'd be a bad joke at best, and observers to be pretty generous to him.
A combination of complacency, lethargy and Woke crap - ‘omg GPT3 might be racist’ - means the west has handed the race to AI to China, and it may already be too late to catch up.
If China dominates AI it dominates the world like no power before it
You should have learnt by now that most people aren’t interested in the really big stories.
We’ve been directly told in the last year that there is definitely ultra tech in our skies and oceans, which either belongs to adversaries of the West or non human intelligence / life forms. And everyone shrugged.
People aren’t going to listen too hard to a senior Pentagon official if he says the US has surrendered technological dominance to China and that a point will be reached (or may already have been reached) when their lead will be insurmountable. Forever.
Cognitive dissonance innit. Much more comfortable to talk about IDS’s majority instead.
Some people are INCREDIBLY resistant to the idea of artificial general intelligence. I have an extremely smart brother who is always open to new ideas but he just won't accept that this - machine intelligence - can ever happen, let alone that it is actually happening right now
Existentially, it frightens him
Or alternatively: your extremely smart brother is actually smarter than you, and realises that everything you rave about is actually fairly sh*t smoke 'n mirrors. and nowhere near 'intelligence' (*). You are staring open-mouthed, dribbling in amazement, as someone performs the three-cup trick.
I remember an august and much-missed member of this board saying that lorry drivers would not be needed due to autonomous driving. That must have been seven or eight years ago now, and we're nowhere near. In fact, we're now suffering from a shortage of drivers. Can you recall him?
The big problem with AI is the money being swilled into the trough. Billions are being spewed at it, and they need to show results. Hence smoke 'n mirrors. Just ask Musky baby where his coast-to-cast drive in a Tesla is - promised five years ago for four years ago. Yet his current tech cannot even detect emergency vehicles...
Machine learning has many uses. But they are limited in scope, and nowhere near a general intelligence. IMO that will require a massive breakthrough in tech, not the stuff we're doing atm.
You'd be amazed at an Eliza produced by drunken first-year undergrads at the West of Scotland Uni...
(*) However you define that.
Yawwwwwwn
See below. In the next ten years we will create a computer that sails through the Turing Test, at that point everything you say here will be rendered irrelevant, whether it is true or not. That *will* be AI, because we will not be able to distinguish it from human intelligence (except that it might be much cleverer and faster)
That's why the Turing Test is such a stroke of genius. It sidesteps all the philosophy and gives you a practical threshold
Actually, we won't.
We'll be able to have a computer that appears - conversationally - to be human. But if you start trying to teach it a new skill, or ask it a puzzle or a riddle, or to explain why a joke is funny, it will fall comically flat.
Can’t we send it to open mic night at The Comedy Store, and tell it to listen to the audience distinguishing between the funny and the not-funny jokes?
A combination of complacency, lethargy and Woke crap - ‘omg GPT3 might be racist’ - means the west has handed the race to AI to China, and it may already be too late to catch up.
If China dominates AI it dominates the world like no power before it
You should have learnt by now that most people aren’t interested in the really big stories.
We’ve been directly told in the last year that there is definitely ultra tech in our skies and oceans, which either belongs to adversaries of the West or non human intelligence / life forms. And everyone shrugged.
People aren’t going to listen too hard to a senior Pentagon official if he says the US has surrendered technological dominance to China and that a point will be reached (or may already have been reached) when their lead will be insurmountable. Forever.
Cognitive dissonance innit. Much more comfortable to talk about IDS’s majority instead.
Some people are INCREDIBLY resistant to the idea of artificial general intelligence. I have an extremely smart brother who is always open to new ideas but he just won't accept that this - machine intelligence - can ever happen, let alone that it is actually happening right now
Existentially, it frightens him
Or alternatively: your extremely smart brother is actually smarter than you, and realises that everything you rave about is actually fairly sh*t smoke 'n mirrors. and nowhere near 'intelligence' (*). You are staring open-mouthed, dribbling in amazement, as someone performs the three-cup trick.
I remember an august and much-missed member of this board saying that lorry drivers would not be needed due to autonomous driving. That must have been seven or eight years ago now, and we're nowhere near. In fact, we're now suffering from a shortage of drivers. Can you recall him?
The big problem with AI is the money being swilled into the trough. Billions are being spewed at it, and they need to show results. Hence smoke 'n mirrors. Just ask Musky baby where his coast-to-cast drive in a Tesla is - promised five years ago for four years ago. Yet his current tech cannot even detect emergency vehicles...
Machine learning has many uses. But they are limited in scope, and nowhere near a general intelligence. IMO that will require a massive breakthrough in tech, not the stuff we're doing atm.
You'd be amazed at an Eliza produced by drunken first-year undergrads at the West of Scotland Uni...
(*) However you define that.
Yawwwwwwn
See below. In the next ten years we will create a computer that sails through the Turing Test, at that point everything you say here will be rendered irrelevant, whether it is true or not. That *will* be AI, because we will not be able to distinguish it from human intelligence (except that it might be much cleverer and faster)
That's why the Turing Test is such a stroke of genius. It sidesteps all the philosophy and gives you a practical threshold
Actually, we won't.
We'll be able to have a computer that appears - conversationally - to be human. But if you start trying to teach it a new skill, or ask it a puzzle or a riddle, or to explain why a joke is funny, it will fall comically flat.
What we mean by AI is pretty important. Once you can programme a machine to make the first few moves in chess sensibly then it is merely a progression in quantity for it to be able to discuss the works of Boethius in Icelandic.
But is it at any level saying it because it thinks it, intuits it, believes it, knows it, has chosen between equally decent alternatives, or alternatively because it has been given enough data to be able to act as if it thinks it? The first is existentially important. The second less so.
The suggestion above that the Turing Test 'sidesteps philosophy', if true, is merely another way of saying it dispenses with the question of whether AI knows anything. Sidestepping reality would be a better term.
The Turing test is given an unduly easy ride. It's merely a stipulation, it's too easy and already been passed in my case - I got a long way into a chat with a chatbot the other day before I asked it Are you human, and it said I don't understand the question. I'd make it harder and stipulate the AI has to sign up as an online-only undergraduate, fool its tutors for three years and get a 2.1 or better at the end of it. And after all that you still haven't got around the self-awareness problem, because we aren't around that with each other yet. I know I am self-aware, I kind of assume you are too because you are like me in many other respects, but I have no hard evidence you are not just a meat computer in a Chinese room. I don't even know you are made of meat, actually, but even if we met the problem wouldn't go away. I don't really see what the evidence would evn consist of.
Funnily enough, they're trying to do exactly that. The Chinese version of GPT3 is enrolling in a university as a student
Erm, isn't it the left of Trump neo-fascism states that will be leaving???
No. There is currently a Democratic President and a Democratic Congress so obviously it will be Trump states, especially in the South that would be most likely to think about seceding.
Remember too every state of the old Confederacy voted for Trump in 2020 except for Virginia and Georgia (and Georgia voted for Trump in 2016)
A combination of complacency, lethargy and Woke crap - ‘omg GPT3 might be racist’ - means the west has handed the race to AI to China, and it may already be too late to catch up.
If China dominates AI it dominates the world like no power before it
You should have learnt by now that most people aren’t interested in the really big stories.
We’ve been directly told in the last year that there is definitely ultra tech in our skies and oceans, which either belongs to adversaries of the West or non human intelligence / life forms. And everyone shrugged.
People aren’t going to listen too hard to a senior Pentagon official if he says the US has surrendered technological dominance to China and that a point will be reached (or may already have been reached) when their lead will be insurmountable. Forever.
Cognitive dissonance innit. Much more comfortable to talk about IDS’s majority instead.
Some people are INCREDIBLY resistant to the idea of artificial general intelligence. I have an extremely smart brother who is always open to new ideas but he just won't accept that this - machine intelligence - can ever happen, let alone that it is actually happening right now
Existentially, it frightens him
Or alternatively: your extremely smart brother is actually smarter than you, and realises that everything you rave about is actually fairly sh*t smoke 'n mirrors. and nowhere near 'intelligence' (*). You are staring open-mouthed, dribbling in amazement, as someone performs the three-cup trick.
I remember an august and much-missed member of this board saying that lorry drivers would not be needed due to autonomous driving. That must have been seven or eight years ago now, and we're nowhere near. In fact, we're now suffering from a shortage of drivers. Can you recall him?
The big problem with AI is the money being swilled into the trough. Billions are being spewed at it, and they need to show results. Hence smoke 'n mirrors. Just ask Musky baby where his coast-to-cast drive in a Tesla is - promised five years ago for four years ago. Yet his current tech cannot even detect emergency vehicles...
Machine learning has many uses. But they are limited in scope, and nowhere near a general intelligence. IMO that will require a massive breakthrough in tech, not the stuff we're doing atm.
You'd be amazed at an Eliza produced by drunken first-year undergrads at the West of Scotland Uni...
(*) However you define that.
Yawwwwwwn
See below. In the next ten years we will create a computer that sails through the Turing Test, at that point everything you say here will be rendered irrelevant, whether it is true or not. That *will* be AI, because we will not be able to distinguish it from human intelligence (except that it might be much cleverer and faster)
That's why the Turing Test is such a stroke of genius. It sidesteps all the philosophy and gives you a practical threshold
I don't think that this will happen. The Turing test isn't that interesting anyway. It's when a computer claims itself to be sentient that the fun starts.
This is amazing. Arguably, this is a computer passing a *kind* of Turing Test. This is Google Duplex booking a haircut, the lady at the salon believes she is talking to a human, the appointment is made, the call ends. Turing Test "passed".
But is it really passed? This guy thinks "no, not really" and I agree with him. However, I also agree with his conclusion:
"There is no way possible that we will not have a general conversational AI in the next 10 years that can speak to any human in any language about every possible topic."
With all due respect, conversation AI is not that hard, it's just giving plausible responses to inputs.
But try explaining set theory to a computer, and then get it to give you examples back.
The difference between *learning plausible responses from a giant corpus* and *learning concepts from conversation* is as wide as the Atlantc.
Now, will we get there? Of course we will. Will it happen in the next ten years? Highly, highly unlikely. Indeed, even assuming exponential intelligence growth, it may well be fifty years away.
But that is how you pass the Turing Test. The computer can only pass the tests given to it. You're shifting the goalposts
And you're also missing the point. If a computer can persuade us it is intelligent - even human - in its interactions with us, then it is, to all intents and purposes, intelligent. That's the point of the Turing Test. Is it ACTUALLY intelligent? Conscious? Thinking? Probably not, but who knows, and Immanuel Kant would struggle to give a definitive answer. And we can no more delve into its wires to seek the answer than we can open up a human brain and locate the place of consciousness
And of course once AI reaches the stage of being indistinguishable from "true" intelligence the answer to all this is, in some senses, immaterial. There will be beings walking or talking amongst us and we will not know if they are human, or not
This is going to have momentous effects, especially online
Up until last year there was a wonderful thing called the Loebner Prize, which was basically the Turing test. And it awarded a prize for the chat bot best able to fool a human they were human.
If you read the conversations, even the ones from 2006/2007, you'll be amazed how good the responses were. Why? Because they were basically doing the same thing as GPT3 - based on the previous five words, what is the most likely sixth word?
That's not a hard problem to solve. Given billions of gigabytes of searchable text, people can easily create systems that give plausible responses. Because all they are really doing is trying to work out what the next word (or sentence) is likely to be.
But even though this is generally considered to be the best (most Turing winning) bot out there, it's still very little better than the bots from 2006/7.
Because the next stage is really hard. GPT3 is not generalised intelligence. Deep Mind is not generalised intelligence. They are really exciting, to be sure, but they don't do what you think they do.
I know what they fucking do. They autocomplete. That's it
But you seem to think that constitutes intelligence.
It doesn't.
Who the fuck knows what truly constitutes intelligence, let alone "self awareness". It is a mystery that has perplexed philosophers for thousands of years, I have a feeling Texas car park software salesman Bobby Smithson has not personally cracked the conundrum which has confounded the greatest human minds since Aristotle
That's the point I've been making for years: 'intelligence' is exceptionally hard to define, and therefore an artificial intelligence is incredibly hard to define. I've heard experts from Cambridge argue over the definition of intelligence, and they're probably still at it.
However, it is possible to say what *isn't* intelligence. A machine putting a square peg into holes in a box at random until it fits in one isn't intelligence. And the ML stuff that you're raving about isn't intelligence either.
"Tory MP James Gray is dropped from role at St John Ambulance 'after saying "They all look the same to me" about Asian ministers Nadhim Zahawi and Sajid Javid' at reception
Tory MP for North Wiltshire James Gray made remark at charity bash Mr Gray, 66, is St John Ambulance commander and was at Parliament reception It was held to recognise 'extraordinary efforts' of volunteers and front-line staff He introduced Mr Zahawi to the stage as the Health Minister but was corrected Mr Gray then was said to comment 'They all look the same to me' to audience He acknowledged it was not appropriate and Mr Zahawi spoke to him privately Last night St John Ambulance said it did not tolerate racism in any form It has now asked him to stand down from all charity activities at once Mr Gray has denied saying remark and insists it was identity mix-up not racism"
It's entirely in character - he created a story just a few weeks ago with a joke about a bomb being sent to another MP, and seems to crop up every few years that way. He doesn't even appear to have the pseudo notoriety of some other perennial backbenchers.
He does, I remember some very senior Tory MPs from a few years ago calling James Gray an absolute copper bottomed shit (and much worse) because of this.
"Tory MP James Gray is dropped from role at St John Ambulance 'after saying "They all look the same to me" about Asian ministers Nadhim Zahawi and Sajid Javid' at reception
Tory MP for North Wiltshire James Gray made remark at charity bash Mr Gray, 66, is St John Ambulance commander and was at Parliament reception It was held to recognise 'extraordinary efforts' of volunteers and front-line staff He introduced Mr Zahawi to the stage as the Health Minister but was corrected Mr Gray then was said to comment 'They all look the same to me' to audience He acknowledged it was not appropriate and Mr Zahawi spoke to him privately Last night St John Ambulance said it did not tolerate racism in any form It has now asked him to stand down from all charity activities at once Mr Gray has denied saying remark and insists it was identity mix-up not racism"
"Tory MP James Gray is dropped from role at St John Ambulance 'after saying "They all look the same to me" about Asian ministers Nadhim Zahawi and Sajid Javid' at reception
Tory MP for North Wiltshire James Gray made remark at charity bash Mr Gray, 66, is St John Ambulance commander and was at Parliament reception It was held to recognise 'extraordinary efforts' of volunteers and front-line staff He introduced Mr Zahawi to the stage as the Health Minister but was corrected Mr Gray then was said to comment 'They all look the same to me' to audience He acknowledged it was not appropriate and Mr Zahawi spoke to him privately Last night St John Ambulance said it did not tolerate racism in any form It has now asked him to stand down from all charity activities at once Mr Gray has denied saying remark and insists it was identity mix-up not racism"
30% of people don't know or care about climate change, 20% of people care and will make sacrifices, 50% want to do something but also drive around and eat burgers on sun holidays.
The point is that we don't really have to make too many sacrifices.
Business Class aside, perhaps.
Completely restructuring the entire world economy to replace cheap and proven sources of energy with expensive and unreliable sources that require huge generation redundancy and unproven mechanisms for energy storage all in a couple of decades. It will be a doddle and of course we'll hardly notice the impact on living standards!
"Tory MP James Gray is dropped from role at St John Ambulance 'after saying "They all look the same to me" about Asian ministers Nadhim Zahawi and Sajid Javid' at reception
Tory MP for North Wiltshire James Gray made remark at charity bash Mr Gray, 66, is St John Ambulance commander and was at Parliament reception It was held to recognise 'extraordinary efforts' of volunteers and front-line staff He introduced Mr Zahawi to the stage as the Health Minister but was corrected Mr Gray then was said to comment 'They all look the same to me' to audience He acknowledged it was not appropriate and Mr Zahawi spoke to him privately Last night St John Ambulance said it did not tolerate racism in any form It has now asked him to stand down from all charity activities at once Mr Gray has denied saying remark and insists it was identity mix-up not racism"
It's entirely in character - he created a story just a few weeks ago with a joke about a bomb being sent to another MP, and seems to crop up every few years that way. He doesn't even appear to have the pseudo notoriety of some other perennial backbenchers.
He does, I remember some very senior Tory MPs from a few years ago calling James Gray an absolute copper bottomed shit (and much worse) because of this.
That's what I mean by in character. But by notoriety I mean there are people like Bill Cash who have always been awkward backbenchers, but who get some attention among the public on a particular issue.
"Tory MP James Gray is dropped from role at St John Ambulance 'after saying "They all look the same to me" about Asian ministers Nadhim Zahawi and Sajid Javid' at reception
Tory MP for North Wiltshire James Gray made remark at charity bash Mr Gray, 66, is St John Ambulance commander and was at Parliament reception It was held to recognise 'extraordinary efforts' of volunteers and front-line staff He introduced Mr Zahawi to the stage as the Health Minister but was corrected Mr Gray then was said to comment 'They all look the same to me' to audience He acknowledged it was not appropriate and Mr Zahawi spoke to him privately Last night St John Ambulance said it did not tolerate racism in any form It has now asked him to stand down from all charity activities at once Mr Gray has denied saying remark and insists it was identity mix-up not racism"
Well, no. To take only the most obvious point, if his eyesight’s so bad he can’t tell the difference between somebody with a beard and somebody without, I don’t want him doing first aid on me anyway.
"Tory MP James Gray is dropped from role at St John Ambulance 'after saying "They all look the same to me" about Asian ministers Nadhim Zahawi and Sajid Javid' at reception
Tory MP for North Wiltshire James Gray made remark at charity bash Mr Gray, 66, is St John Ambulance commander and was at Parliament reception It was held to recognise 'extraordinary efforts' of volunteers and front-line staff He introduced Mr Zahawi to the stage as the Health Minister but was corrected Mr Gray then was said to comment 'They all look the same to me' to audience He acknowledged it was not appropriate and Mr Zahawi spoke to him privately Last night St John Ambulance said it did not tolerate racism in any form It has now asked him to stand down from all charity activities at once Mr Gray has denied saying remark and insists it was identity mix-up not racism"
"Tory MP James Gray is dropped from role at St John Ambulance 'after saying "They all look the same to me" about Asian ministers Nadhim Zahawi and Sajid Javid' at reception
Tory MP for North Wiltshire James Gray made remark at charity bash Mr Gray, 66, is St John Ambulance commander and was at Parliament reception It was held to recognise 'extraordinary efforts' of volunteers and front-line staff He introduced Mr Zahawi to the stage as the Health Minister but was corrected Mr Gray then was said to comment 'They all look the same to me' to audience He acknowledged it was not appropriate and Mr Zahawi spoke to him privately Last night St John Ambulance said it did not tolerate racism in any form It has now asked him to stand down from all charity activities at once Mr Gray has denied saying remark and insists it was identity mix-up not racism"
It is very odd however you look at it because Nadhim Zahawi looks nothing like The Saj. Even the most extreme racist would have no trouble telling them apart. It tells of impaired judgement first to get them mixed up and secondly to remark they all look the same. Gray should consider his position not because he is racist but because he is clearly not thinking straight.
Since, as you note, they don't look remotely alike, I could believe it was a genuine confusion on his part and the comment was an attempt at a joke. But that requires him to lean in hard with a 'I'm stupid' defence, since it'd be a bad joke at best, and observers to be pretty generous to him.
Gray can probably get away with a profuse apology and announcing he will retire at the next election. As he is 66, that is probably true anyway.
"Tory MP James Gray is dropped from role at St John Ambulance 'after saying "They all look the same to me" about Asian ministers Nadhim Zahawi and Sajid Javid' at reception
Tory MP for North Wiltshire James Gray made remark at charity bash Mr Gray, 66, is St John Ambulance commander and was at Parliament reception It was held to recognise 'extraordinary efforts' of volunteers and front-line staff He introduced Mr Zahawi to the stage as the Health Minister but was corrected Mr Gray then was said to comment 'They all look the same to me' to audience He acknowledged it was not appropriate and Mr Zahawi spoke to him privately Last night St John Ambulance said it did not tolerate racism in any form It has now asked him to stand down from all charity activities at once Mr Gray has denied saying remark and insists it was identity mix-up not racism"
A combination of complacency, lethargy and Woke crap - ‘omg GPT3 might be racist’ - means the west has handed the race to AI to China, and it may already be too late to catch up.
If China dominates AI it dominates the world like no power before it
You should have learnt by now that most people aren’t interested in the really big stories.
We’ve been directly told in the last year that there is definitely ultra tech in our skies and oceans, which either belongs to adversaries of the West or non human intelligence / life forms. And everyone shrugged.
People aren’t going to listen too hard to a senior Pentagon official if he says the US has surrendered technological dominance to China and that a point will be reached (or may already have been reached) when their lead will be insurmountable. Forever.
Cognitive dissonance innit. Much more comfortable to talk about IDS’s majority instead.
Some people are INCREDIBLY resistant to the idea of artificial general intelligence. I have an extremely smart brother who is always open to new ideas but he just won't accept that this - machine intelligence - can ever happen, let alone that it is actually happening right now
Existentially, it frightens him
Or alternatively: your extremely smart brother is actually smarter than you, and realises that everything you rave about is actually fairly sh*t smoke 'n mirrors. and nowhere near 'intelligence' (*). You are staring open-mouthed, dribbling in amazement, as someone performs the three-cup trick.
I remember an august and much-missed member of this board saying that lorry drivers would not be needed due to autonomous driving. That must have been seven or eight years ago now, and we're nowhere near. In fact, we're now suffering from a shortage of drivers. Can you recall him?
The big problem with AI is the money being swilled into the trough. Billions are being spewed at it, and they need to show results. Hence smoke 'n mirrors. Just ask Musky baby where his coast-to-cast drive in a Tesla is - promised five years ago for four years ago. Yet his current tech cannot even detect emergency vehicles...
Machine learning has many uses. But they are limited in scope, and nowhere near a general intelligence. IMO that will require a massive breakthrough in tech, not the stuff we're doing atm.
You'd be amazed at an Eliza produced by drunken first-year undergrads at the West of Scotland Uni...
(*) However you define that.
Yawwwwwwn
See below. In the next ten years we will create a computer that sails through the Turing Test, at that point everything you say here will be rendered irrelevant, whether it is true or not. That *will* be AI, because we will not be able to distinguish it from human intelligence (except that it might be much cleverer and faster)
That's why the Turing Test is such a stroke of genius. It sidesteps all the philosophy and gives you a practical threshold
I don't think that this will happen. The Turing test isn't that interesting anyway. It's when a computer claims itself to be sentient that the fun starts.
This is amazing. Arguably, this is a computer passing a *kind* of Turing Test. This is Google Duplex booking a haircut, the lady at the salon believes she is talking to a human, the appointment is made, the call ends. Turing Test "passed".
But is it really passed? This guy thinks "no, not really" and I agree with him. However, I also agree with his conclusion:
"There is no way possible that we will not have a general conversational AI in the next 10 years that can speak to any human in any language about every possible topic."
With all due respect, conversation AI is not that hard, it's just giving plausible responses to inputs.
But try explaining set theory to a computer, and then get it to give you examples back.
The difference between *learning plausible responses from a giant corpus* and *learning concepts from conversation* is as wide as the Atlantc.
Now, will we get there? Of course we will. Will it happen in the next ten years? Highly, highly unlikely. Indeed, even assuming exponential intelligence growth, it may well be fifty years away.
But that is how you pass the Turing Test. The computer can only pass the tests given to it. You're shifting the goalposts
And you're also missing the point. If a computer can persuade us it is intelligent - even human - in its interactions with us, then it is, to all intents and purposes, intelligent. That's the point of the Turing Test. Is it ACTUALLY intelligent? Conscious? Thinking? Probably not, but who knows, and Immanuel Kant would struggle to give a definitive answer. And we can no more delve into its wires to seek the answer than we can open up a human brain and locate the place of consciousness
And of course once AI reaches the stage of being indistinguishable from "true" intelligence the answer to all this is, in some senses, immaterial. There will be beings walking or talking amongst us and we will not know if they are human, or not
This is going to have momentous effects, especially online
Up until last year there was a wonderful thing called the Loebner Prize, which was basically the Turing test. And it awarded a prize for the chat bot best able to fool a human they were human.
If you read the conversations, even the ones from 2006/2007, you'll be amazed how good the responses were. Why? Because they were basically doing the same thing as GPT3 - based on the previous five words, what is the most likely sixth word?
That's not a hard problem to solve. Given billions of gigabytes of searchable text, people can easily create systems that give plausible responses. Because all they are really doing is trying to work out what the next word (or sentence) is likely to be.
But even though this is generally considered to be the best (most Turing winning) bot out there, it's still very little better than the bots from 2006/7.
Because the next stage is really hard. GPT3 is not generalised intelligence. Deep Mind is not generalised intelligence. They are really exciting, to be sure, but they don't do what you think they do.
I know what they fucking do. They autocomplete. That's it
But you seem to think that constitutes intelligence.
It doesn't.
Who the fuck knows what truly constitutes intelligence, let alone "self awareness". It is a mystery that has perplexed philosophers for thousands of years, I have a feeling Texas car park software salesman Bobby Smithson has not personally cracked the conundrum which has confounded the greatest human minds since Aristotle
That's the point I've been making for years: 'intelligence' is exceptionally hard to define, and therefore an artificial intelligence is incredibly hard to define. I've heard experts from Cambridge argue over the definition of intelligence, and they're probably still at it.
However, it is possible to say what *isn't* intelligence. A machine putting a square peg into holes in a box at random until it fits in one isn't intelligence. And the ML stuff that you're raving about isn't intelligence either.
Head::Desk
Which is why, of course, Alan Turing invented his test, to sidestep the impossibly complex and contentious arguments around intelligence, consciousness, Theory of Mind, and all that.
Instead of trying to define intelligence and locate consciousness, you admit that it's indefinable "but you know it when you see it". Like pornography
The Turing Test is the "know it when you see it" approach: to intelligence. If the computer can convince humans it is human and intelligent, it is intelligent. This is the whole fucking point of it, tho that seems to have escaped quite a few PB-ers, probably because they are as about as smart as a ZX Spectrum
"Tory MP James Gray is dropped from role at St John Ambulance 'after saying "They all look the same to me" about Asian ministers Nadhim Zahawi and Sajid Javid' at reception
Tory MP for North Wiltshire James Gray made remark at charity bash Mr Gray, 66, is St John Ambulance commander and was at Parliament reception It was held to recognise 'extraordinary efforts' of volunteers and front-line staff He introduced Mr Zahawi to the stage as the Health Minister but was corrected Mr Gray then was said to comment 'They all look the same to me' to audience He acknowledged it was not appropriate and Mr Zahawi spoke to him privately Last night St John Ambulance said it did not tolerate racism in any form It has now asked him to stand down from all charity activities at once Mr Gray has denied saying remark and insists it was identity mix-up not racism"
"Tory MP James Gray is dropped from role at St John Ambulance 'after saying "They all look the same to me" about Asian ministers Nadhim Zahawi and Sajid Javid' at reception
Tory MP for North Wiltshire James Gray made remark at charity bash Mr Gray, 66, is St John Ambulance commander and was at Parliament reception It was held to recognise 'extraordinary efforts' of volunteers and front-line staff He introduced Mr Zahawi to the stage as the Health Minister but was corrected Mr Gray then was said to comment 'They all look the same to me' to audience He acknowledged it was not appropriate and Mr Zahawi spoke to him privately Last night St John Ambulance said it did not tolerate racism in any form It has now asked him to stand down from all charity activities at once Mr Gray has denied saying remark and insists it was identity mix-up not racism"
It is very odd however you look at it because Nadhim Zahawi looks nothing like The Saj. Even the most extreme racist would have no trouble telling them apart. It tells of impaired judgement first to get them mixed up and secondly to remark they all look the same. Gray should consider his position not because he is racist but because he is clearly not thinking straight.
Since, as you note, they don't look remotely alike, I could believe it was a genuine confusion on his part and the comment was an attempt at a joke. But that requires him to lean in hard with a 'I'm stupid' defence, since it'd be a bad joke at best, and observers to be pretty generous to him.
Gray can probably get away with a profuse apology and announcing he will retire at the next election. As he is 66, that is probably true anyway.
I doubt he needs to even go that far. It'll die down, and 66 is not old for an MP.
That said, his seat is proposed to be chopped up and joined with parts of Gloucestershire.
Erm, isn't it the left of Trump neo-fascism states that will be leaving???
No. There is currently a Democratic President and a Democratic Congress so obviously it will be Trump states, especially in the South that would be most likely to think about seceding.
Remember too every state of the old Confederacy voted for Trump in 2020 except for Virginia and Georgia (and Georgia voted for Trump in 2016)
I can see Putin's bots rubbing their hands at that one.
A combination of complacency, lethargy and Woke crap - ‘omg GPT3 might be racist’ - means the west has handed the race to AI to China, and it may already be too late to catch up.
If China dominates AI it dominates the world like no power before it
You should have learnt by now that most people aren’t interested in the really big stories.
We’ve been directly told in the last year that there is definitely ultra tech in our skies and oceans, which either belongs to adversaries of the West or non human intelligence / life forms. And everyone shrugged.
People aren’t going to listen too hard to a senior Pentagon official if he says the US has surrendered technological dominance to China and that a point will be reached (or may already have been reached) when their lead will be insurmountable. Forever.
Cognitive dissonance innit. Much more comfortable to talk about IDS’s majority instead.
Some people are INCREDIBLY resistant to the idea of artificial general intelligence. I have an extremely smart brother who is always open to new ideas but he just won't accept that this - machine intelligence - can ever happen, let alone that it is actually happening right now
Existentially, it frightens him
Or alternatively: your extremely smart brother is actually smarter than you, and realises that everything you rave about is actually fairly sh*t smoke 'n mirrors. and nowhere near 'intelligence' (*). You are staring open-mouthed, dribbling in amazement, as someone performs the three-cup trick.
I remember an august and much-missed member of this board saying that lorry drivers would not be needed due to autonomous driving. That must have been seven or eight years ago now, and we're nowhere near. In fact, we're now suffering from a shortage of drivers. Can you recall him?
The big problem with AI is the money being swilled into the trough. Billions are being spewed at it, and they need to show results. Hence smoke 'n mirrors. Just ask Musky baby where his coast-to-cast drive in a Tesla is - promised five years ago for four years ago. Yet his current tech cannot even detect emergency vehicles...
Machine learning has many uses. But they are limited in scope, and nowhere near a general intelligence. IMO that will require a massive breakthrough in tech, not the stuff we're doing atm.
You'd be amazed at an Eliza produced by drunken first-year undergrads at the West of Scotland Uni...
(*) However you define that.
Yawwwwwwn
See below. In the next ten years we will create a computer that sails through the Turing Test, at that point everything you say here will be rendered irrelevant, whether it is true or not. That *will* be AI, because we will not be able to distinguish it from human intelligence (except that it might be much cleverer and faster)
That's why the Turing Test is such a stroke of genius. It sidesteps all the philosophy and gives you a practical threshold
I don't think that this will happen. The Turing test isn't that interesting anyway. It's when a computer claims itself to be sentient that the fun starts.
This is amazing. Arguably, this is a computer passing a *kind* of Turing Test. This is Google Duplex booking a haircut, the lady at the salon believes she is talking to a human, the appointment is made, the call ends. Turing Test "passed".
But is it really passed? This guy thinks "no, not really" and I agree with him. However, I also agree with his conclusion:
"There is no way possible that we will not have a general conversational AI in the next 10 years that can speak to any human in any language about every possible topic."
With all due respect, conversation AI is not that hard, it's just giving plausible responses to inputs.
But try explaining set theory to a computer, and then get it to give you examples back.
The difference between *learning plausible responses from a giant corpus* and *learning concepts from conversation* is as wide as the Atlantc.
Now, will we get there? Of course we will. Will it happen in the next ten years? Highly, highly unlikely. Indeed, even assuming exponential intelligence growth, it may well be fifty years away.
But that is how you pass the Turing Test. The computer can only pass the tests given to it. You're shifting the goalposts
And you're also missing the point. If a computer can persuade us it is intelligent - even human - in its interactions with us, then it is, to all intents and purposes, intelligent. That's the point of the Turing Test. Is it ACTUALLY intelligent? Conscious? Thinking? Probably not, but who knows, and Immanuel Kant would struggle to give a definitive answer. And we can no more delve into its wires to seek the answer than we can open up a human brain and locate the place of consciousness
And of course once AI reaches the stage of being indistinguishable from "true" intelligence the answer to all this is, in some senses, immaterial. There will be beings walking or talking amongst us and we will not know if they are human, or not
This is going to have momentous effects, especially online
Up until last year there was a wonderful thing called the Loebner Prize, which was basically the Turing test. And it awarded a prize for the chat bot best able to fool a human they were human.
If you read the conversations, even the ones from 2006/2007, you'll be amazed how good the responses were. Why? Because they were basically doing the same thing as GPT3 - based on the previous five words, what is the most likely sixth word?
That's not a hard problem to solve. Given billions of gigabytes of searchable text, people can easily create systems that give plausible responses. Because all they are really doing is trying to work out what the next word (or sentence) is likely to be.
But even though this is generally considered to be the best (most Turing winning) bot out there, it's still very little better than the bots from 2006/7.
Because the next stage is really hard. GPT3 is not generalised intelligence. Deep Mind is not generalised intelligence. They are really exciting, to be sure, but they don't do what you think they do.
I know what they fucking do. They autocomplete. That's it
But you seem to think that constitutes intelligence.
It doesn't.
Who the fuck knows what truly constitutes intelligence, let alone "self awareness". It is a mystery that has perplexed philosophers for thousands of years, I have a feeling Texas car park software salesman Bobby Smithson has not personally cracked the conundrum which has confounded the greatest human minds since Aristotle
A combination of complacency, lethargy and Woke crap - ‘omg GPT3 might be racist’ - means the west has handed the race to AI to China, and it may already be too late to catch up.
If China dominates AI it dominates the world like no power before it
You should have learnt by now that most people aren’t interested in the really big stories.
We’ve been directly told in the last year that there is definitely ultra tech in our skies and oceans, which either belongs to adversaries of the West or non human intelligence / life forms. And everyone shrugged.
People aren’t going to listen too hard to a senior Pentagon official if he says the US has surrendered technological dominance to China and that a point will be reached (or may already have been reached) when their lead will be insurmountable. Forever.
Cognitive dissonance innit. Much more comfortable to talk about IDS’s majority instead.
Some people are INCREDIBLY resistant to the idea of artificial general intelligence. I have an extremely smart brother who is always open to new ideas but he just won't accept that this - machine intelligence - can ever happen, let alone that it is actually happening right now
Existentially, it frightens him
Or alternatively: your extremely smart brother is actually smarter than you, and realises that everything you rave about is actually fairly sh*t smoke 'n mirrors. and nowhere near 'intelligence' (*). You are staring open-mouthed, dribbling in amazement, as someone performs the three-cup trick.
I remember an august and much-missed member of this board saying that lorry drivers would not be needed due to autonomous driving. That must have been seven or eight years ago now, and we're nowhere near. In fact, we're now suffering from a shortage of drivers. Can you recall him?
The big problem with AI is the money being swilled into the trough. Billions are being spewed at it, and they need to show results. Hence smoke 'n mirrors. Just ask Musky baby where his coast-to-cast drive in a Tesla is - promised five years ago for four years ago. Yet his current tech cannot even detect emergency vehicles...
Machine learning has many uses. But they are limited in scope, and nowhere near a general intelligence. IMO that will require a massive breakthrough in tech, not the stuff we're doing atm.
You'd be amazed at an Eliza produced by drunken first-year undergrads at the West of Scotland Uni...
(*) However you define that.
Yawwwwwwn
See below. In the next ten years we will create a computer that sails through the Turing Test, at that point everything you say here will be rendered irrelevant, whether it is true or not. That *will* be AI, because we will not be able to distinguish it from human intelligence (except that it might be much cleverer and faster)
That's why the Turing Test is such a stroke of genius. It sidesteps all the philosophy and gives you a practical threshold
Actually, we won't.
We'll be able to have a computer that appears - conversationally - to be human. But if you start trying to teach it a new skill, or ask it a puzzle or a riddle, or to explain why a joke is funny, it will fall comically flat.
What we mean by AI is pretty important. Once you can programme a machine to make the first few moves in chess sensibly then it is merely a progression in quantity for it to be able to discuss the works of Boethius in Icelandic.
But is it at any level saying it because it thinks it, intuits it, believes it, knows it, has chosen between equally decent alternatives, or alternatively because it has been given enough data to be able to act as if it thinks it? The first is existentially important. The second less so.
The suggestion above that the Turing Test 'sidesteps philosophy', if true, is merely another way of saying it dispenses with the question of whether AI knows anything. Sidestepping reality would be a better term.
The Turing test is given an unduly easy ride. It's merely a stipulation, it's too easy and already been passed in my case - I got a long way into a chat with a chatbot the other day before I asked it Are you human, and it said I don't understand the question. I'd make it harder and stipulate the AI has to sign up as an online-only undergraduate, fool its tutors for three years and get a 2.1 or better at the end of it. And after all that you still haven't got around the self-awareness problem, because we aren't around that with each other yet. I know I am self-aware, I kind of assume you are too because you are like me in many other respects, but I have no hard evidence you are not just a meat computer in a Chinese room. I don't even know you are made of meat, actually, but even if we met the problem wouldn't go away. I don't really see what the evidence would evn consist of.
Funnily enough, they're trying to do exactly that. The Chinese version of GPT3 is enrolling in a university as a student
Just look at that article and see the warning signs. Amusingly, the picture is of an extremely attractive woman (and it is always a woman).
I'm also wary when people talk about the size of the datasets going into it. Large numbers sound impressive. Datasets are useful for ML; less so for real AI. And GIGO is a massive factor with datasets. Where are the datasets coming from? Have they been verified? Are they fit for this purpose?
Also, the claims are from a researcher, and there is zero independent verification of the work.
Until we get such, it is all just smoke 'n mirrors.
A combination of complacency, lethargy and Woke crap - ‘omg GPT3 might be racist’ - means the west has handed the race to AI to China, and it may already be too late to catch up.
If China dominates AI it dominates the world like no power before it
You should have learnt by now that most people aren’t interested in the really big stories.
We’ve been directly told in the last year that there is definitely ultra tech in our skies and oceans, which either belongs to adversaries of the West or non human intelligence / life forms. And everyone shrugged.
People aren’t going to listen too hard to a senior Pentagon official if he says the US has surrendered technological dominance to China and that a point will be reached (or may already have been reached) when their lead will be insurmountable. Forever.
Cognitive dissonance innit. Much more comfortable to talk about IDS’s majority instead.
Some people are INCREDIBLY resistant to the idea of artificial general intelligence. I have an extremely smart brother who is always open to new ideas but he just won't accept that this - machine intelligence - can ever happen, let alone that it is actually happening right now
Existentially, it frightens him
Or alternatively: your extremely smart brother is actually smarter than you, and realises that everything you rave about is actually fairly sh*t smoke 'n mirrors. and nowhere near 'intelligence' (*). You are staring open-mouthed, dribbling in amazement, as someone performs the three-cup trick.
I remember an august and much-missed member of this board saying that lorry drivers would not be needed due to autonomous driving. That must have been seven or eight years ago now, and we're nowhere near. In fact, we're now suffering from a shortage of drivers. Can you recall him?
The big problem with AI is the money being swilled into the trough. Billions are being spewed at it, and they need to show results. Hence smoke 'n mirrors. Just ask Musky baby where his coast-to-cast drive in a Tesla is - promised five years ago for four years ago. Yet his current tech cannot even detect emergency vehicles...
Machine learning has many uses. But they are limited in scope, and nowhere near a general intelligence. IMO that will require a massive breakthrough in tech, not the stuff we're doing atm.
You'd be amazed at an Eliza produced by drunken first-year undergrads at the West of Scotland Uni...
(*) However you define that.
Yawwwwwwn
See below. In the next ten years we will create a computer that sails through the Turing Test, at that point everything you say here will be rendered irrelevant, whether it is true or not. That *will* be AI, because we will not be able to distinguish it from human intelligence (except that it might be much cleverer and faster)
That's why the Turing Test is such a stroke of genius. It sidesteps all the philosophy and gives you a practical threshold
I don't think that this will happen. The Turing test isn't that interesting anyway. It's when a computer claims itself to be sentient that the fun starts.
This is amazing. Arguably, this is a computer passing a *kind* of Turing Test. This is Google Duplex booking a haircut, the lady at the salon believes she is talking to a human, the appointment is made, the call ends. Turing Test "passed".
But is it really passed? This guy thinks "no, not really" and I agree with him. However, I also agree with his conclusion:
"There is no way possible that we will not have a general conversational AI in the next 10 years that can speak to any human in any language about every possible topic."
With all due respect, conversation AI is not that hard, it's just giving plausible responses to inputs.
But try explaining set theory to a computer, and then get it to give you examples back.
The difference between *learning plausible responses from a giant corpus* and *learning concepts from conversation* is as wide as the Atlantc.
Now, will we get there? Of course we will. Will it happen in the next ten years? Highly, highly unlikely. Indeed, even assuming exponential intelligence growth, it may well be fifty years away.
But that is how you pass the Turing Test. The computer can only pass the tests given to it. You're shifting the goalposts
And you're also missing the point. If a computer can persuade us it is intelligent - even human - in its interactions with us, then it is, to all intents and purposes, intelligent. That's the point of the Turing Test. Is it ACTUALLY intelligent? Conscious? Thinking? Probably not, but who knows, and Immanuel Kant would struggle to give a definitive answer. And we can no more delve into its wires to seek the answer than we can open up a human brain and locate the place of consciousness
And of course once AI reaches the stage of being indistinguishable from "true" intelligence the answer to all this is, in some senses, immaterial. There will be beings walking or talking amongst us and we will not know if they are human, or not
This is going to have momentous effects, especially online
Up until last year there was a wonderful thing called the Loebner Prize, which was basically the Turing test. And it awarded a prize for the chat bot best able to fool a human they were human.
If you read the conversations, even the ones from 2006/2007, you'll be amazed how good the responses were. Why? Because they were basically doing the same thing as GPT3 - based on the previous five words, what is the most likely sixth word?
That's not a hard problem to solve. Given billions of gigabytes of searchable text, people can easily create systems that give plausible responses. Because all they are really doing is trying to work out what the next word (or sentence) is likely to be.
But even though this is generally considered to be the best (most Turing winning) bot out there, it's still very little better than the bots from 2006/7.
Because the next stage is really hard. GPT3 is not generalised intelligence. Deep Mind is not generalised intelligence. They are really exciting, to be sure, but they don't do what you think they do.
I know what they fucking do. They autocomplete. That's it
But you seem to think that constitutes intelligence.
It doesn't.
Who the fuck knows what truly constitutes intelligence, let alone "self awareness". It is a mystery that has perplexed philosophers for thousands of years, I have a feeling Texas car park software salesman Bobby Smithson has not personally cracked the conundrum which has confounded the greatest human minds since Aristotle
A combination of complacency, lethargy and Woke crap - ‘omg GPT3 might be racist’ - means the west has handed the race to AI to China, and it may already be too late to catch up.
If China dominates AI it dominates the world like no power before it
You should have learnt by now that most people aren’t interested in the really big stories.
We’ve been directly told in the last year that there is definitely ultra tech in our skies and oceans, which either belongs to adversaries of the West or non human intelligence / life forms. And everyone shrugged.
People aren’t going to listen too hard to a senior Pentagon official if he says the US has surrendered technological dominance to China and that a point will be reached (or may already have been reached) when their lead will be insurmountable. Forever.
Cognitive dissonance innit. Much more comfortable to talk about IDS’s majority instead.
Some people are INCREDIBLY resistant to the idea of artificial general intelligence. I have an extremely smart brother who is always open to new ideas but he just won't accept that this - machine intelligence - can ever happen, let alone that it is actually happening right now
Existentially, it frightens him
Or alternatively: your extremely smart brother is actually smarter than you, and realises that everything you rave about is actually fairly sh*t smoke 'n mirrors. and nowhere near 'intelligence' (*). You are staring open-mouthed, dribbling in amazement, as someone performs the three-cup trick.
I remember an august and much-missed member of this board saying that lorry drivers would not be needed due to autonomous driving. That must have been seven or eight years ago now, and we're nowhere near. In fact, we're now suffering from a shortage of drivers. Can you recall him?
The big problem with AI is the money being swilled into the trough. Billions are being spewed at it, and they need to show results. Hence smoke 'n mirrors. Just ask Musky baby where his coast-to-cast drive in a Tesla is - promised five years ago for four years ago. Yet his current tech cannot even detect emergency vehicles...
Machine learning has many uses. But they are limited in scope, and nowhere near a general intelligence. IMO that will require a massive breakthrough in tech, not the stuff we're doing atm.
You'd be amazed at an Eliza produced by drunken first-year undergrads at the West of Scotland Uni...
(*) However you define that.
Yawwwwwwn
See below. In the next ten years we will create a computer that sails through the Turing Test, at that point everything you say here will be rendered irrelevant, whether it is true or not. That *will* be AI, because we will not be able to distinguish it from human intelligence (except that it might be much cleverer and faster)
That's why the Turing Test is such a stroke of genius. It sidesteps all the philosophy and gives you a practical threshold
I don't think that this will happen. The Turing test isn't that interesting anyway. It's when a computer claims itself to be sentient that the fun starts.
This is amazing. Arguably, this is a computer passing a *kind* of Turing Test. This is Google Duplex booking a haircut, the lady at the salon believes she is talking to a human, the appointment is made, the call ends. Turing Test "passed".
But is it really passed? This guy thinks "no, not really" and I agree with him. However, I also agree with his conclusion:
"There is no way possible that we will not have a general conversational AI in the next 10 years that can speak to any human in any language about every possible topic."
With all due respect, conversation AI is not that hard, it's just giving plausible responses to inputs.
But try explaining set theory to a computer, and then get it to give you examples back.
The difference between *learning plausible responses from a giant corpus* and *learning concepts from conversation* is as wide as the Atlantc.
Now, will we get there? Of course we will. Will it happen in the next ten years? Highly, highly unlikely. Indeed, even assuming exponential intelligence growth, it may well be fifty years away.
But that is how you pass the Turing Test. The computer can only pass the tests given to it. You're shifting the goalposts
And you're also missing the point. If a computer can persuade us it is intelligent - even human - in its interactions with us, then it is, to all intents and purposes, intelligent. That's the point of the Turing Test. Is it ACTUALLY intelligent? Conscious? Thinking? Probably not, but who knows, and Immanuel Kant would struggle to give a definitive answer. And we can no more delve into its wires to seek the answer than we can open up a human brain and locate the place of consciousness
And of course once AI reaches the stage of being indistinguishable from "true" intelligence the answer to all this is, in some senses, immaterial. There will be beings walking or talking amongst us and we will not know if they are human, or not
This is going to have momentous effects, especially online
Up until last year there was a wonderful thing called the Loebner Prize, which was basically the Turing test. And it awarded a prize for the chat bot best able to fool a human they were human.
If you read the conversations, even the ones from 2006/2007, you'll be amazed how good the responses were. Why? Because they were basically doing the same thing as GPT3 - based on the previous five words, what is the most likely sixth word?
That's not a hard problem to solve. Given billions of gigabytes of searchable text, people can easily create systems that give plausible responses. Because all they are really doing is trying to work out what the next word (or sentence) is likely to be.
But even though this is generally considered to be the best (most Turing winning) bot out there, it's still very little better than the bots from 2006/7.
Because the next stage is really hard. GPT3 is not generalised intelligence. Deep Mind is not generalised intelligence. They are really exciting, to be sure, but they don't do what you think they do.
I know what they fucking do. They autocomplete. That's it
But you seem to think that constitutes intelligence.
It doesn't.
Who the fuck knows what truly constitutes intelligence, let alone "self awareness". It is a mystery that has perplexed philosophers for thousands of years, I have a feeling Texas car park software salesman Bobby Smithson has not personally cracked the conundrum which has confounded the greatest human minds since Aristotle
That's the point I've been making for years: 'intelligence' is exceptionally hard to define, and therefore an artificial intelligence is incredibly hard to define. I've heard experts from Cambridge argue over the definition of intelligence, and they're probably still at it.
However, it is possible to say what *isn't* intelligence. A machine putting a square peg into holes in a box at random until it fits in one isn't intelligence. And the ML stuff that you're raving about isn't intelligence either.
Head::Desk
Which is why, of course, Alan Turing invented his test, to sidestep the impossibly complex and contentious arguments around intelligence, consciousness, Theory of Mind, and all that.
Instead of trying to define intelligence and locate consciousness, you admit that it's indefinable "but you know it when you see it". Like pornography
The Turing Test is the "know it when you see it" approach: to intelligence. If the computer can convince humans it is human and intelligent, it is intelligent. This is the whole fucking point of it, tho that seems to have escaped quite a few PB-ers, probably because they are as about as smart as a ZX Spectrum
But as others have pointed out (including, I believe, Turing himself), the Turing Test is very limited.
Also, Turing did not envisage the amount of data that could be used to create a system that could fake intelligence.
Why does this constant over-hyping of AI matter? Because pretending dumb systems are intelligent can have massive real-world effects:
"Tory MP James Gray is dropped from role at St John Ambulance 'after saying "They all look the same to me" about Asian ministers Nadhim Zahawi and Sajid Javid' at reception
Tory MP for North Wiltshire James Gray made remark at charity bash Mr Gray, 66, is St John Ambulance commander and was at Parliament reception It was held to recognise 'extraordinary efforts' of volunteers and front-line staff He introduced Mr Zahawi to the stage as the Health Minister but was corrected Mr Gray then was said to comment 'They all look the same to me' to audience He acknowledged it was not appropriate and Mr Zahawi spoke to him privately Last night St John Ambulance said it did not tolerate racism in any form It has now asked him to stand down from all charity activities at once Mr Gray has denied saying remark and insists it was identity mix-up not racism"
It is very odd however you look at it because Nadhim Zahawi looks nothing like The Saj. Even the most extreme racist would have no trouble telling them apart. It tells of impaired judgement first to get them mixed up and secondly to remark they all look the same. Gray should consider his position not because he is racist but because he is clearly not thinking straight.
Since, as you note, they don't look remotely alike, I could believe it was a genuine confusion on his part and the comment was an attempt at a joke. But that requires him to lean in hard with a 'I'm stupid' defence, since it'd be a bad joke at best, and observers to be pretty generous to him.
Gray can probably get away with a profuse apology and announcing he will retire at the next election. As he is 66, that is probably true anyway.
If the report is correct, Gray said: "They all look the same to me".
The word "all" is the give-away, indicating a racist joke; there were only two of them. If he'd simply said "They (both) look the same to me", an innocent (albeit stupid because they don't look alike) error is plausible.
And it was only last week that he 'joked' about delivering a bomb to Anneliese Dodds' office.
A combination of complacency, lethargy and Woke crap - ‘omg GPT3 might be racist’ - means the west has handed the race to AI to China, and it may already be too late to catch up.
If China dominates AI it dominates the world like no power before it
You should have learnt by now that most people aren’t interested in the really big stories.
We’ve been directly told in the last year that there is definitely ultra tech in our skies and oceans, which either belongs to adversaries of the West or non human intelligence / life forms. And everyone shrugged.
People aren’t going to listen too hard to a senior Pentagon official if he says the US has surrendered technological dominance to China and that a point will be reached (or may already have been reached) when their lead will be insurmountable. Forever.
Cognitive dissonance innit. Much more comfortable to talk about IDS’s majority instead.
Some people are INCREDIBLY resistant to the idea of artificial general intelligence. I have an extremely smart brother who is always open to new ideas but he just won't accept that this - machine intelligence - can ever happen, let alone that it is actually happening right now
Existentially, it frightens him
Or alternatively: your extremely smart brother is actually smarter than you, and realises that everything you rave about is actually fairly sh*t smoke 'n mirrors. and nowhere near 'intelligence' (*). You are staring open-mouthed, dribbling in amazement, as someone performs the three-cup trick.
I remember an august and much-missed member of this board saying that lorry drivers would not be needed due to autonomous driving. That must have been seven or eight years ago now, and we're nowhere near. In fact, we're now suffering from a shortage of drivers. Can you recall him?
The big problem with AI is the money being swilled into the trough. Billions are being spewed at it, and they need to show results. Hence smoke 'n mirrors. Just ask Musky baby where his coast-to-cast drive in a Tesla is - promised five years ago for four years ago. Yet his current tech cannot even detect emergency vehicles...
Machine learning has many uses. But they are limited in scope, and nowhere near a general intelligence. IMO that will require a massive breakthrough in tech, not the stuff we're doing atm.
You'd be amazed at an Eliza produced by drunken first-year undergrads at the West of Scotland Uni...
(*) However you define that.
Yawwwwwwn
See below. In the next ten years we will create a computer that sails through the Turing Test, at that point everything you say here will be rendered irrelevant, whether it is true or not. That *will* be AI, because we will not be able to distinguish it from human intelligence (except that it might be much cleverer and faster)
That's why the Turing Test is such a stroke of genius. It sidesteps all the philosophy and gives you a practical threshold
I don't think that this will happen. The Turing test isn't that interesting anyway. It's when a computer claims itself to be sentient that the fun starts.
This is amazing. Arguably, this is a computer passing a *kind* of Turing Test. This is Google Duplex booking a haircut, the lady at the salon believes she is talking to a human, the appointment is made, the call ends. Turing Test "passed".
But is it really passed? This guy thinks "no, not really" and I agree with him. However, I also agree with his conclusion:
"There is no way possible that we will not have a general conversational AI in the next 10 years that can speak to any human in any language about every possible topic."
With all due respect, conversation AI is not that hard, it's just giving plausible responses to inputs.
But try explaining set theory to a computer, and then get it to give you examples back.
The difference between *learning plausible responses from a giant corpus* and *learning concepts from conversation* is as wide as the Atlantc.
Now, will we get there? Of course we will. Will it happen in the next ten years? Highly, highly unlikely. Indeed, even assuming exponential intelligence growth, it may well be fifty years away.
But that is how you pass the Turing Test. The computer can only pass the tests given to it. You're shifting the goalposts
And you're also missing the point. If a computer can persuade us it is intelligent - even human - in its interactions with us, then it is, to all intents and purposes, intelligent. That's the point of the Turing Test. Is it ACTUALLY intelligent? Conscious? Thinking? Probably not, but who knows, and Immanuel Kant would struggle to give a definitive answer. And we can no more delve into its wires to seek the answer than we can open up a human brain and locate the place of consciousness
And of course once AI reaches the stage of being indistinguishable from "true" intelligence the answer to all this is, in some senses, immaterial. There will be beings walking or talking amongst us and we will not know if they are human, or not
This is going to have momentous effects, especially online
Up until last year there was a wonderful thing called the Loebner Prize, which was basically the Turing test. And it awarded a prize for the chat bot best able to fool a human they were human.
If you read the conversations, even the ones from 2006/2007, you'll be amazed how good the responses were. Why? Because they were basically doing the same thing as GPT3 - based on the previous five words, what is the most likely sixth word?
That's not a hard problem to solve. Given billions of gigabytes of searchable text, people can easily create systems that give plausible responses. Because all they are really doing is trying to work out what the next word (or sentence) is likely to be.
But even though this is generally considered to be the best (most Turing winning) bot out there, it's still very little better than the bots from 2006/7.
Because the next stage is really hard. GPT3 is not generalised intelligence. Deep Mind is not generalised intelligence. They are really exciting, to be sure, but they don't do what you think they do.
I know what they fucking do. They autocomplete. That's it
But you seem to think that constitutes intelligence.
It doesn't.
Who the fuck knows what truly constitutes intelligence, let alone "self awareness". It is a mystery that has perplexed philosophers for thousands of years, I have a feeling Texas car park software salesman Bobby Smithson has not personally cracked the conundrum which has confounded the greatest human minds since Aristotle
That's the point I've been making for years: 'intelligence' is exceptionally hard to define, and therefore an artificial intelligence is incredibly hard to define. I've heard experts from Cambridge argue over the definition of intelligence, and they're probably still at it.
However, it is possible to say what *isn't* intelligence. A machine putting a square peg into holes in a box at random until it fits in one isn't intelligence. And the ML stuff that you're raving about isn't intelligence either.
Head::Desk
Which is why, of course, Alan Turing invented his test, to sidestep the impossibly complex and contentious arguments around intelligence, consciousness, Theory of Mind, and all that.
Instead of trying to define intelligence and locate consciousness, you admit that it's indefinable "but you know it when you see it". Like pornography
The Turing Test is the "know it when you see it" approach: to intelligence. If the computer can convince humans it is human and intelligent, it is intelligent. This is the whole fucking point of it, tho that seems to have escaped quite a few PB-ers, probably because they are as about as smart as a ZX Spectrum
The philosophy of the Turing Test is fascinating.
However you basically don't seem to be aware of it at all.
If you have a tightly defined specification people can optimise the program they write to meet the specification that otherwise would not in any way be seen to be intelligent.
This is where the Chinese Room wankers actually have a point.
Imagine a computer program that is simply a list of all plausible things a human might say in a general conversation that is attempting to determine if the agent is human or not. For every entry in that list there is an output sentence to be repeated in response.
Boom. The list is unimaginably huge, completely rote and obviously in no way intelligent. Yet would pass the vast majority of converstations as human,
"Tory MP James Gray is dropped from role at St John Ambulance 'after saying "They all look the same to me" about Asian ministers Nadhim Zahawi and Sajid Javid' at reception
Tory MP for North Wiltshire James Gray made remark at charity bash Mr Gray, 66, is St John Ambulance commander and was at Parliament reception It was held to recognise 'extraordinary efforts' of volunteers and front-line staff He introduced Mr Zahawi to the stage as the Health Minister but was corrected Mr Gray then was said to comment 'They all look the same to me' to audience He acknowledged it was not appropriate and Mr Zahawi spoke to him privately Last night St John Ambulance said it did not tolerate racism in any form It has now asked him to stand down from all charity activities at once Mr Gray has denied saying remark and insists it was identity mix-up not racism"
It is very odd however you look at it because Nadhim Zahawi looks nothing like The Saj. Even the most extreme racist would have no trouble telling them apart. It tells of impaired judgement first to get them mixed up and secondly to remark they all look the same. Gray should consider his position not because he is racist but because he is clearly not thinking straight.
Since, as you note, they don't look remotely alike, I could believe it was a genuine confusion on his part and the comment was an attempt at a joke. But that requires him to lean in hard with a 'I'm stupid' defence, since it'd be a bad joke at best, and observers to be pretty generous to him.
Gray can probably get away with a profuse apology and announcing he will retire at the next election. As he is 66, that is probably true anyway.
If the report is correct, Gray said: "They all look the same to me".
The word "all" is the give-away, indicating a racist joke. If he'd simply said "They look the same to me", he could well get away with it.
And it was only last week that he 'joked' about delivering a bomb to Anneliese Dodds' office.
If he’d said ‘they look the same to me’ it should surely have resulted in the immediate loss of his driving licence.
If I had to guess I'd say we're on course for 2-2.5C of warming at its highest point.
That won't be fun and will cause a lot of extreme weather, migration and political disruption (remember: we are already at 1.2C of warming now) and the loss of quite a few marginal species but we won't become "extinct".
Far from it.
That would probably be enough warming to trigger a complete melt of the Greenland Ice Sheet, which would have state-ending consequences for many nations - how does Egypt survive without the Nile Delta?
IIRC a combined total melt of the Greenland and West Antarctic ice caps would be enough to raise sea levels by about 10m (the East Antarctic ice cap, which is the real biggie, would add another 60m but it isn't thought to be under immediate threat.) However, that said, this process is likely to take many hundreds and possibly thousands of years to complete, so people living in most of the potentially affected areas today have nothing to worry about.
Er, what? Joke, surely? God intended America to be promised land for European Christians (% AGREE): 30% all 52% wh evangelical Prot 53% Republican 52% trust Fox News 67% trust far right news 65% QAnon believer @PRRIpoll
No time to die - great movie. Second best of the Craig run after Casino Royale. Really enjoyed it, my wife did as well and she's not a huge fan of the series. Lea Seydoux looks and talks uncannily like my wife which was a bit jarring for me personally.
I'd rate it as 5th best Bond movie.
1. GoldenEye 2. Casino Royale 3. Thunderball 4. Live and Let Die 5. No Time to Die
A combination of complacency, lethargy and Woke crap - ‘omg GPT3 might be racist’ - means the west has handed the race to AI to China, and it may already be too late to catch up.
If China dominates AI it dominates the world like no power before it
You should have learnt by now that most people aren’t interested in the really big stories.
We’ve been directly told in the last year that there is definitely ultra tech in our skies and oceans, which either belongs to adversaries of the West or non human intelligence / life forms. And everyone shrugged.
People aren’t going to listen too hard to a senior Pentagon official if he says the US has surrendered technological dominance to China and that a point will be reached (or may already have been reached) when their lead will be insurmountable. Forever.
Cognitive dissonance innit. Much more comfortable to talk about IDS’s majority instead.
Some people are INCREDIBLY resistant to the idea of artificial general intelligence. I have an extremely smart brother who is always open to new ideas but he just won't accept that this - machine intelligence - can ever happen, let alone that it is actually happening right now
Existentially, it frightens him
Or alternatively: your extremely smart brother is actually smarter than you, and realises that everything you rave about is actually fairly sh*t smoke 'n mirrors. and nowhere near 'intelligence' (*). You are staring open-mouthed, dribbling in amazement, as someone performs the three-cup trick.
I remember an august and much-missed member of this board saying that lorry drivers would not be needed due to autonomous driving. That must have been seven or eight years ago now, and we're nowhere near. In fact, we're now suffering from a shortage of drivers. Can you recall him?
The big problem with AI is the money being swilled into the trough. Billions are being spewed at it, and they need to show results. Hence smoke 'n mirrors. Just ask Musky baby where his coast-to-cast drive in a Tesla is - promised five years ago for four years ago. Yet his current tech cannot even detect emergency vehicles...
Machine learning has many uses. But they are limited in scope, and nowhere near a general intelligence. IMO that will require a massive breakthrough in tech, not the stuff we're doing atm.
You'd be amazed at an Eliza produced by drunken first-year undergrads at the West of Scotland Uni...
(*) However you define that.
Yawwwwwwn
See below. In the next ten years we will create a computer that sails through the Turing Test, at that point everything you say here will be rendered irrelevant, whether it is true or not. That *will* be AI, because we will not be able to distinguish it from human intelligence (except that it might be much cleverer and faster)
That's why the Turing Test is such a stroke of genius. It sidesteps all the philosophy and gives you a practical threshold
I don't think that this will happen. The Turing test isn't that interesting anyway. It's when a computer claims itself to be sentient that the fun starts.
This is amazing. Arguably, this is a computer passing a *kind* of Turing Test. This is Google Duplex booking a haircut, the lady at the salon believes she is talking to a human, the appointment is made, the call ends. Turing Test "passed".
But is it really passed? This guy thinks "no, not really" and I agree with him. However, I also agree with his conclusion:
"There is no way possible that we will not have a general conversational AI in the next 10 years that can speak to any human in any language about every possible topic."
With all due respect, conversation AI is not that hard, it's just giving plausible responses to inputs.
But try explaining set theory to a computer, and then get it to give you examples back.
The difference between *learning plausible responses from a giant corpus* and *learning concepts from conversation* is as wide as the Atlantc.
Now, will we get there? Of course we will. Will it happen in the next ten years? Highly, highly unlikely. Indeed, even assuming exponential intelligence growth, it may well be fifty years away.
But that is how you pass the Turing Test. The computer can only pass the tests given to it. You're shifting the goalposts
And you're also missing the point. If a computer can persuade us it is intelligent - even human - in its interactions with us, then it is, to all intents and purposes, intelligent. That's the point of the Turing Test. Is it ACTUALLY intelligent? Conscious? Thinking? Probably not, but who knows, and Immanuel Kant would struggle to give a definitive answer. And we can no more delve into its wires to seek the answer than we can open up a human brain and locate the place of consciousness
And of course once AI reaches the stage of being indistinguishable from "true" intelligence the answer to all this is, in some senses, immaterial. There will be beings walking or talking amongst us and we will not know if they are human, or not
This is going to have momentous effects, especially online
Up until last year there was a wonderful thing called the Loebner Prize, which was basically the Turing test. And it awarded a prize for the chat bot best able to fool a human they were human.
If you read the conversations, even the ones from 2006/2007, you'll be amazed how good the responses were. Why? Because they were basically doing the same thing as GPT3 - based on the previous five words, what is the most likely sixth word?
That's not a hard problem to solve. Given billions of gigabytes of searchable text, people can easily create systems that give plausible responses. Because all they are really doing is trying to work out what the next word (or sentence) is likely to be.
But even though this is generally considered to be the best (most Turing winning) bot out there, it's still very little better than the bots from 2006/7.
Because the next stage is really hard. GPT3 is not generalised intelligence. Deep Mind is not generalised intelligence. They are really exciting, to be sure, but they don't do what you think they do.
I know what they fucking do. They autocomplete. That's it
But you seem to think that constitutes intelligence.
It doesn't.
Who the fuck knows what truly constitutes intelligence, let alone "self awareness". It is a mystery that has perplexed philosophers for thousands of years, I have a feeling Texas car park software salesman Bobby Smithson has not personally cracked the conundrum which has confounded the greatest human minds since Aristotle
That's the point I've been making for years: 'intelligence' is exceptionally hard to define, and therefore an artificial intelligence is incredibly hard to define. I've heard experts from Cambridge argue over the definition of intelligence, and they're probably still at it.
However, it is possible to say what *isn't* intelligence. A machine putting a square peg into holes in a box at random until it fits in one isn't intelligence. And the ML stuff that you're raving about isn't intelligence either.
Head::Desk
Which is why, of course, Alan Turing invented his test, to sidestep the impossibly complex and contentious arguments around intelligence, consciousness, Theory of Mind, and all that.
Instead of trying to define intelligence and locate consciousness, you admit that it's indefinable "but you know it when you see it". Like pornography
The Turing Test is the "know it when you see it" approach: to intelligence. If the computer can convince humans it is human and intelligent, it is intelligent. This is the whole fucking point of it, tho that seems to have escaped quite a few PB-ers, probably because they are as about as smart as a ZX Spectrum
The philosophy of the Turing Test is fascinating.
However you basically don't seem to be aware of it at all.
If you have a tightly defined specification people can optimise the program they write to meet the specification that otherwise would not in any way be seen to be intelligent.
This is where the Chinese Room wankers actually have a point.
Imagine a computer program that is simply a list of all plausible things a human might say in a general conversation that is attempting to determine if the agent is human or not. For every entry in that list there is an output sentence to be repeated in response.
Boom. The list is unimaginably huge, completely rote and obviously in no way intelligent. Yet would pass the vast majority of converstations as human,
The Turing Test is exactly as I describe, and useful because of it:
"The power and appeal of the Turing test derives from its simplicity. The philosophy of mind, psychology, and modern neuroscience have been unable to provide definitions of "intelligence" and "thinking" that are sufficiently precise and general to be applied to machines. Without such definitions, the central questions of the philosophy of artificial intelligence cannot be answered. The Turing test, even if imperfect, at least provides something that can actually be measured. As such, it is a pragmatic attempt to answer a difficult philosophical question."
Turing himself, in his original paper, proposed his test so as to avoid the impossible enigma of "consciousness"
"I do not wish to give the impression that I think there is no mystery about consciousness. There is, for instance, something of a paradox connected with any attempt to localise it. But I do not think these mysteries necessarily need to be solved before we can answer the question with which we are concerned in this paper."
When I am bored I sometimes play a game with PB. I go to a longer comment (avoiding the name at the top) and I read the first 2 or 3 sentences. And then I test myself: I see if I can guess the identity of the commenter from the opinion, syntax, vocabulary, style
It is surprisingly easy, we all have a style. What is more surprising is how robotic and repetitive some commenters are, such that you can not only guess the identity of the commenter, but you can predict what they will say next, after those first 2 sentences, sometimes down to the precise word.
Two extreme examples are Kinabalu on the left, and HYUFD on the right. No offence guys, but I suggest you are actually bots on Russia's SputnikGPT-3 in Chelyabinsk, autocompleting your comments following the prompt of a prior comment - as that is how GPT3 works. It is basically "autocomplete on crack"
That raises a further question, one I have mentioned before. What if ALL intelligence is just autocomplete? We think we have original thoughts, ideas, concepts, but maybe all of us - not just the twin droids kinabalu and HYUFD - are just a bunch of algorithms, responding as we must?
If all intelligence is just autocomplete, then AI is already here, and it is called GPT3, and it only going to get more intelligent
Here's a fascinating essay exploring that exact same idea that I had last year. Or, I should say, that idea I thought i had, in reality it was just me autocompleting the new reality of Natural Language Programming
The most fun posters are those whose views on a subject are unpredictable.
Hat tip to @IshmaelZ, who is reasonably inscrutable.
Bit of fun, see how many people can get of these:
(1) "100% disagreed. If the price of human body parts goes down demand for them will rise until equilibrium is found. Do you deny this? Yes/No."
(2) "Mr kinabalu, there’s a reason why Usain Bolt’s entrance into the 4th form girls’ egg and spoon race would be deemed by most to be inappropriate."
(3) "High intensity burns, preferably at the gym and in close proximity to a “htg” doing unfeasible contortions on a mat, that’s my ying to the yang of an otherwise indulgent lifestyle. When I’ve accomplished 30 secs of that, and got my breath back, I feel splendid."
(4) "Afternoon team, good debating from everyone today, anybody know where I can buy a new bike, mine’s crocked. TIA."
(5) "PB Tories might be able to articulate something borderline intelligible every so often if they weren’t always gagging on Johnson’s cock."
(6) "Had it with men in dresses and wanker scientists who talk wank the whole time. I’m going to Switzerland."
(7) "I was at a dinner in Islington the other evening and had terrific fun demonstrating to the people there how the green belt is an example of the Payroll being kept at the gates of the castle by their lords and masters."
(8) "You must think I button up the back you half-witted cretin. Jog on."
(9) "I know this won’t be popular on here but my feeling is that the more the Dems bang on about the Capitol “riots” and the attempted “coup” (lol) the more people will notice their risible double standards and look forward to voting for Trump next time."
(10) "Can't see why woke is sapping the moral fibre of the West? Cognitive bias, pure and simple."
(11) "Sky News. Boris says everything is coming up roses. Nothing from Labour."
(12) "Crumpets for tea is possible if Starmer is PM in a hung parliament being propped up by Sturgeon and the SNP otherwise no chance because Boris has said it can’t be for another 40 years."
Apols to those missing. Could have done loads more.
Are these actual quotes, or your impersonations of PB folk? Either way, hope you'll provide the key at some stage after we've all had fun guessing.
Er, what? Joke, surely? God intended America to be promised land for European Christians (% AGREE): 30% all 52% wh evangelical Prot 53% Republican 52% trust Fox News 67% trust far right news 65% QAnon believer @PRRIpoll
"Tory MP James Gray is dropped from role at St John Ambulance 'after saying "They all look the same to me" about Asian ministers Nadhim Zahawi and Sajid Javid' at reception
Tory MP for North Wiltshire James Gray made remark at charity bash Mr Gray, 66, is St John Ambulance commander and was at Parliament reception It was held to recognise 'extraordinary efforts' of volunteers and front-line staff He introduced Mr Zahawi to the stage as the Health Minister but was corrected Mr Gray then was said to comment 'They all look the same to me' to audience He acknowledged it was not appropriate and Mr Zahawi spoke to him privately Last night St John Ambulance said it did not tolerate racism in any form It has now asked him to stand down from all charity activities at once Mr Gray has denied saying remark and insists it was identity mix-up not racism"
It is very odd however you look at it because Nadhim Zahawi looks nothing like The Saj. Even the most extreme racist would have no trouble telling them apart. It tells of impaired judgement first to get them mixed up and secondly to remark they all look the same. Gray should consider his position not because he is racist but because he is clearly not thinking straight.
Although I did think that there was a certain similarity between Martin Houghton-Brown(?) the CEO of St John’s Ambulance and Nadhim Zahawi in the photo of the 3 of them together
A combination of complacency, lethargy and Woke crap - ‘omg GPT3 might be racist’ - means the west has handed the race to AI to China, and it may already be too late to catch up.
If China dominates AI it dominates the world like no power before it
You should have learnt by now that most people aren’t interested in the really big stories.
We’ve been directly told in the last year that there is definitely ultra tech in our skies and oceans, which either belongs to adversaries of the West or non human intelligence / life forms. And everyone shrugged.
People aren’t going to listen too hard to a senior Pentagon official if he says the US has surrendered technological dominance to China and that a point will be reached (or may already have been reached) when their lead will be insurmountable. Forever.
Cognitive dissonance innit. Much more comfortable to talk about IDS’s majority instead.
Some people are INCREDIBLY resistant to the idea of artificial general intelligence. I have an extremely smart brother who is always open to new ideas but he just won't accept that this - machine intelligence - can ever happen, let alone that it is actually happening right now
Existentially, it frightens him
Or alternatively: your extremely smart brother is actually smarter than you, and realises that everything you rave about is actually fairly sh*t smoke 'n mirrors. and nowhere near 'intelligence' (*). You are staring open-mouthed, dribbling in amazement, as someone performs the three-cup trick.
I remember an august and much-missed member of this board saying that lorry drivers would not be needed due to autonomous driving. That must have been seven or eight years ago now, and we're nowhere near. In fact, we're now suffering from a shortage of drivers. Can you recall him?
The big problem with AI is the money being swilled into the trough. Billions are being spewed at it, and they need to show results. Hence smoke 'n mirrors. Just ask Musky baby where his coast-to-cast drive in a Tesla is - promised five years ago for four years ago. Yet his current tech cannot even detect emergency vehicles...
Machine learning has many uses. But they are limited in scope, and nowhere near a general intelligence. IMO that will require a massive breakthrough in tech, not the stuff we're doing atm.
You'd be amazed at an Eliza produced by drunken first-year undergrads at the West of Scotland Uni...
(*) However you define that.
Yawwwwwwn
See below. In the next ten years we will create a computer that sails through the Turing Test, at that point everything you say here will be rendered irrelevant, whether it is true or not. That *will* be AI, because we will not be able to distinguish it from human intelligence (except that it might be much cleverer and faster)
That's why the Turing Test is such a stroke of genius. It sidesteps all the philosophy and gives you a practical threshold
I don't think that this will happen. The Turing test isn't that interesting anyway. It's when a computer claims itself to be sentient that the fun starts.
This is amazing. Arguably, this is a computer passing a *kind* of Turing Test. This is Google Duplex booking a haircut, the lady at the salon believes she is talking to a human, the appointment is made, the call ends. Turing Test "passed".
But is it really passed? This guy thinks "no, not really" and I agree with him. However, I also agree with his conclusion:
"There is no way possible that we will not have a general conversational AI in the next 10 years that can speak to any human in any language about every possible topic."
With all due respect, conversation AI is not that hard, it's just giving plausible responses to inputs.
But try explaining set theory to a computer, and then get it to give you examples back.
The difference between *learning plausible responses from a giant corpus* and *learning concepts from conversation* is as wide as the Atlantc.
Now, will we get there? Of course we will. Will it happen in the next ten years? Highly, highly unlikely. Indeed, even assuming exponential intelligence growth, it may well be fifty years away.
But that is how you pass the Turing Test. The computer can only pass the tests given to it. You're shifting the goalposts
And you're also missing the point. If a computer can persuade us it is intelligent - even human - in its interactions with us, then it is, to all intents and purposes, intelligent. That's the point of the Turing Test. Is it ACTUALLY intelligent? Conscious? Thinking? Probably not, but who knows, and Immanuel Kant would struggle to give a definitive answer. And we can no more delve into its wires to seek the answer than we can open up a human brain and locate the place of consciousness
And of course once AI reaches the stage of being indistinguishable from "true" intelligence the answer to all this is, in some senses, immaterial. There will be beings walking or talking amongst us and we will not know if they are human, or not
This is going to have momentous effects, especially online
Up until last year there was a wonderful thing called the Loebner Prize, which was basically the Turing test. And it awarded a prize for the chat bot best able to fool a human they were human.
If you read the conversations, even the ones from 2006/2007, you'll be amazed how good the responses were. Why? Because they were basically doing the same thing as GPT3 - based on the previous five words, what is the most likely sixth word?
That's not a hard problem to solve. Given billions of gigabytes of searchable text, people can easily create systems that give plausible responses. Because all they are really doing is trying to work out what the next word (or sentence) is likely to be.
But even though this is generally considered to be the best (most Turing winning) bot out there, it's still very little better than the bots from 2006/7.
Because the next stage is really hard. GPT3 is not generalised intelligence. Deep Mind is not generalised intelligence. They are really exciting, to be sure, but they don't do what you think they do.
I know what they fucking do. They autocomplete. That's it
But you seem to think that constitutes intelligence.
It doesn't.
Who the fuck knows what truly constitutes intelligence, let alone "self awareness". It is a mystery that has perplexed philosophers for thousands of years, I have a feeling Texas car park software salesman Bobby Smithson has not personally cracked the conundrum which has confounded the greatest human minds since Aristotle
That's the point I've been making for years: 'intelligence' is exceptionally hard to define, and therefore an artificial intelligence is incredibly hard to define. I've heard experts from Cambridge argue over the definition of intelligence, and they're probably still at it.
However, it is possible to say what *isn't* intelligence. A machine putting a square peg into holes in a box at random until it fits in one isn't intelligence. And the ML stuff that you're raving about isn't intelligence either.
Head::Desk
Which is why, of course, Alan Turing invented his test, to sidestep the impossibly complex and contentious arguments around intelligence, consciousness, Theory of Mind, and all that.
Instead of trying to define intelligence and locate consciousness, you admit that it's indefinable "but you know it when you see it". Like pornography
The Turing Test is the "know it when you see it" approach: to intelligence. If the computer can convince humans it is human and intelligent, it is intelligent. This is the whole fucking point of it, tho that seems to have escaped quite a few PB-ers, probably because they are as about as smart as a ZX Spectrum
The philosophy of the Turing Test is fascinating.
However you basically don't seem to be aware of it at all.
If you have a tightly defined specification people can optimise the program they write to meet the specification that otherwise would not in any way be seen to be intelligent.
This is where the Chinese Room wankers actually have a point.
Imagine a computer program that is simply a list of all plausible things a human might say in a general conversation that is attempting to determine if the agent is human or not. For every entry in that list there is an output sentence to be repeated in response.
Boom. The list is unimaginably huge, completely rote and obviously in no way intelligent. Yet would pass the vast majority of converstations as human,
The Turing Test is exactly as I describe, and useful because of it:
"The power and appeal of the Turing test derives from its simplicity. The philosophy of mind, psychology, and modern neuroscience have been unable to provide definitions of "intelligence" and "thinking" that are sufficiently precise and general to be applied to machines. Without such definitions, the central questions of the philosophy of artificial intelligence cannot be answered. The Turing test, even if imperfect, at least provides something that can actually be measured. As such, it is a pragmatic attempt to answer a difficult philosophical question."
Turing himself, in his original paper, proposed his test so as to avoid the impossible enigma of "consciousness"
"I do not wish to give the impression that I think there is no mystery about consciousness. There is, for instance, something of a paradox connected with any attempt to localise it. But I do not think these mysteries necessarily need to be solved before we can answer the question with which we are concerned in this paper."
Almost entirely off topic my son has his first Oxford Union debate this week, that this house has no confidence in the government. A certain JRM is appearing for the government as is Geoffrey Cox QC. Should be fun.
O/T a majority believing that there will be more opportunities than costs in attempting net zero doesn't make it so. When the gigantic costs start to become apparent and electricity blackouts start the opinion polls might shift slightly.
The truth is hundreds of millions of people live in coastal cities vulnerable to sea level rise and enhanced storms - will the current Thames Barrier adequately protect London in 20 years?
I remember seeing a video of a talk by Francis Pryor (archaeologist, prehistoric fens specialist) where he predicted in passing that we'll end up allowing the fens to be reflooded within the next 50 years, because the alternative will be allowing full strength North Sea storm surges to carry on round the coast and flood London. He could be completely wrong, of course, but it it ever happens it would be a pretty dramatic change to the landscape around here...
Sitting in Starbucks being exposed to an entire album of some K-Pop girl band. Sounds like someone has shoved the complete works of Lady Gaga, Britney Spears, Beyonce and Pitbull into some machine learning programme and this is what was spat out.
Almost entirely off topic my son has his first Oxford Union debate this week, that this house has no confidence in the government. A certain JRM is appearing for the government as is Geoffrey Cox QC. Should be fun.
No time to die - great movie. Second best of the Craig run after Casino Royale. Really enjoyed it, my wife did as well and she's not a huge fan of the series. Lea Seydoux looks and talks uncannily like my wife which was a bit jarring for me personally.
I'd rate it as 5th best Bond movie.
1. GoldenEye 2. Casino Royale 3. Thunderball 4. Live and Let Die 5. No Time to Die
Would be my list.
Wait.
You have Live and Let Die on there, but no Goldfinger?
No time to die - great movie. Second best of the Craig run after Casino Royale. Really enjoyed it, my wife did as well and she's not a huge fan of the series. Lea Seydoux looks and talks uncannily like my wife which was a bit jarring for me personally.
I'd rate it as 5th best Bond movie.
1. GoldenEye 2. Casino Royale 3. Thunderball 4. Live and Let Die 5. No Time to Die
Would be my list.
LOL! I know the feeling. Can’t watch anything with Jennifer Paige without wondering what my wife is doing!
No time to die - great movie. Second best of the Craig run after Casino Royale. Really enjoyed it, my wife did as well and she's not a huge fan of the series. Lea Seydoux looks and talks uncannily like my wife which was a bit jarring for me personally.
I'd rate it as 5th best Bond movie.
1. GoldenEye 2. Casino Royale 3. Thunderball 4. Live and Let Die 5. No Time to Die
Would be my list.
Wait.
You have Live and Let Die on there, but no Goldfinger?
That's probably within the next few tbf, I really enjoyed Live and Let Die. I know a lot of people didn't.
"Tory MP James Gray is dropped from role at St John Ambulance 'after saying "They all look the same to me" about Asian ministers Nadhim Zahawi and Sajid Javid' at reception
Tory MP for North Wiltshire James Gray made remark at charity bash Mr Gray, 66, is St John Ambulance commander and was at Parliament reception It was held to recognise 'extraordinary efforts' of volunteers and front-line staff He introduced Mr Zahawi to the stage as the Health Minister but was corrected Mr Gray then was said to comment 'They all look the same to me' to audience He acknowledged it was not appropriate and Mr Zahawi spoke to him privately Last night St John Ambulance said it did not tolerate racism in any form It has now asked him to stand down from all charity activities at once Mr Gray has denied saying remark and insists it was identity mix-up not racism"
No time to die - great movie. Second best of the Craig run after Casino Royale. Really enjoyed it, my wife did as well and she's not a huge fan of the series. Lea Seydoux looks and talks uncannily like my wife which was a bit jarring for me personally.
I'd rate it as 5th best Bond movie.
1. GoldenEye 2. Casino Royale 3. Thunderball 4. Live and Let Die 5. No Time to Die
Would be my list.
Wait.
You have Live and Let Die on there, but no Goldfinger?
No time to die - great movie. Second best of the Craig run after Casino Royale. Really enjoyed it, my wife did as well and she's not a huge fan of the series. Lea Seydoux looks and talks uncannily like my wife which was a bit jarring for me personally.
I'd rate it as 5th best Bond movie.
1. GoldenEye 2. Casino Royale 3. Thunderball 4. Live and Let Die 5. No Time to Die
Would be my list.
Wait.
You have Live and Let Die on there, but no Goldfinger?
No time to die - great movie. Second best of the Craig run after Casino Royale. Really enjoyed it, my wife did as well and she's not a huge fan of the series. Lea Seydoux looks and talks uncannily like my wife which was a bit jarring for me personally.
I'd rate it as 5th best Bond movie.
1. GoldenEye 2. Casino Royale 3. Thunderball 4. Live and Let Die 5. No Time to Die
Would be my list.
Wait.
You have Live and Let Die on there, but no Goldfinger?
That's probably within the next few tbf, I really enjoyed Live and Let Die. I know a lot of people didn't.
I really enjoyed it as a kid, but now think it's among the weaker Bond offerings (although I like the theme tune).
I would have
1. Casino Royale 2. Goldfinger 3. Thunderball 4. Goldeneye 5. The Living Daylights (which I think is a much underrated movie)
No time to die - great movie. Second best of the Craig run after Casino Royale. Really enjoyed it, my wife did as well and she's not a huge fan of the series. Lea Seydoux looks and talks uncannily like my wife which was a bit jarring for me personally.
I'd rate it as 5th best Bond movie.
1. GoldenEye 2. Casino Royale 3. Thunderball 4. Live and Let Die 5. No Time to Die
Would be my list.
Wait.
You have Live and Let Die on there, but no Goldfinger?
And where's On Her Majesty's Secret Service?!
You may have all the time in the world to debate this. I don’t.
No time to die - great movie. Second best of the Craig run after Casino Royale. Really enjoyed it, my wife did as well and she's not a huge fan of the series. Lea Seydoux looks and talks uncannily like my wife which was a bit jarring for me personally.
I'd rate it as 5th best Bond movie.
1. GoldenEye 2. Casino Royale 3. Thunderball 4. Live and Let Die 5. No Time to Die
Would be my list.
Wait.
You have Live and Let Die on there, but no Goldfinger?
No time to die - great movie. Second best of the Craig run after Casino Royale. Really enjoyed it, my wife did as well and she's not a huge fan of the series. Lea Seydoux looks and talks uncannily like my wife which was a bit jarring for me personally.
I'd rate it as 5th best Bond movie.
1. GoldenEye 2. Casino Royale 3. Thunderball 4. Live and Let Die 5. No Time to Die
Would be my list.
Wait.
You have Live and Let Die on there, but no Goldfinger?
Almost entirely off topic my son has his first Oxford Union debate this week, that this house has no confidence in the government. A certain JRM is appearing for the government as is Geoffrey Cox QC. Should be fun.
Mass debate, is it?
Mass audience. Nick Thomas-Symonds MP (who?) for the motion along with Layla Moran. Started 15 minutes ago.
"Tory MP James Gray is dropped from role at St John Ambulance 'after saying "They all look the same to me" about Asian ministers Nadhim Zahawi and Sajid Javid' at reception
Tory MP for North Wiltshire James Gray made remark at charity bash Mr Gray, 66, is St John Ambulance commander and was at Parliament reception It was held to recognise 'extraordinary efforts' of volunteers and front-line staff He introduced Mr Zahawi to the stage as the Health Minister but was corrected Mr Gray then was said to comment 'They all look the same to me' to audience He acknowledged it was not appropriate and Mr Zahawi spoke to him privately Last night St John Ambulance said it did not tolerate racism in any form It has now asked him to stand down from all charity activities at once Mr Gray has denied saying remark and insists it was identity mix-up not racism"
It is very odd however you look at it because Nadhim Zahawi looks nothing like The Saj. Even the most extreme racist would have no trouble telling them apart. It tells of impaired judgement first to get them mixed up and secondly to remark they all look the same. Gray should consider his position not because he is racist but because he is clearly not thinking straight.
That would be a great criterion, taking us straight to hung Parliament
Think he hasn't really had a proper look at them either. "Young people", he says. Most of them seem to me to be middle-aged middle-class pillocks.
The one I know about a few miles from me is a retiree millionaire in a horrible-to-heat stone farmhouse, who rents out not-very-green holiday barn conversions for up to £2.5k a week.
Prince of Wales has said world leaders gathering at the Cop26 summit should take ambitious action on climate change rather than “just talk”, and take notice of how “despairing” many young people are about the issue.
Charles said he understood why climate campaign groups such as Extinction Rebellion stage protests and block roads, but suggested they should take a less disruptive approach.
“I totally understand the frustration,” he said in an interview with the BBC.
“But it isn’t helpful, I don’t think, to do it in a way that alienates people … The difficulty is, how do you direct that frustration in a way that is more constructive rather than destructive?”
He added: “The point is, people should really notice how despairing so many young are.”
Like crooked company directors, the only thing that will make them take notice is time in a prison cell.
NEW: people obsess over vaccine uptake stats, eagerly comparing one country to others to see which has jabbed the highest share of its population, but what if I told you many — perhaps most — of those stats are wrong?
Time for a thread on bad Covid data and how it can cost lives
A combination of complacency, lethargy and Woke crap - ‘omg GPT3 might be racist’ - means the west has handed the race to AI to China, and it may already be too late to catch up.
If China dominates AI it dominates the world like no power before it
You should have learnt by now that most people aren’t interested in the really big stories.
We’ve been directly told in the last year that there is definitely ultra tech in our skies and oceans, which either belongs to adversaries of the West or non human intelligence / life forms. And everyone shrugged.
People aren’t going to listen too hard to a senior Pentagon official if he says the US has surrendered technological dominance to China and that a point will be reached (or may already have been reached) when their lead will be insurmountable. Forever.
Cognitive dissonance innit. Much more comfortable to talk about IDS’s majority instead.
Some people are INCREDIBLY resistant to the idea of artificial general intelligence. I have an extremely smart brother who is always open to new ideas but he just won't accept that this - machine intelligence - can ever happen, let alone that it is actually happening right now
Existentially, it frightens him
Or alternativelby drunken first-year undergrads at the West of Scotland Uni...
(*) However you define that.
Yawwwwwwn
See below. In the next ten years we will create a computer that sails through the Turing Test, at that point everything you say here will be rendered irrelevant, whether it is true or not. That *will* be AI, because we will not be able to distinguish it from human intelligence (except that it might be much cleverer and faster)
That's why the Turing Test is such a stroke of genius. It sidesteps all the philosophy and gives you a practical threshold
I don't think that this will happen. The Turing test isn't that interesting anyway. It's when a computer claims itself to be sentient that the fun starts.
This is amazing. Arguably, this is a computer passing a *kind* of Turing Test. This is Google Duplex booking a haircut, the lady at the salon believes she is talking to a human, the appointment is made, the call ends. Turing Test "passed".
But is it really passed? This guy thinks "no, not really" and I agree with him. However, I also agree with his conclusion:
"There is no way possible that we will not have a general conversational AI in the next 10 years that can speak to any human in any language about every possible topic."
With all due respect, conversation AI is not that hard, it's just giving plausible responses to inputs.
But try explaining set theory to a computer, and then get it to give you examples back.
The difference between *learning plausible responses from a giant corpus* and *learning concepts from conversation* is as wide as the Atlantc.
Now, will we get there? Of course we will. Will it happen in the next ten years? Highly, highly unlikely. Indeed, even assuming exponential intelligence growth, it may well be fifty years away.
But that is hopecially online
Up until last year there was a wonderful thing called the Loebner Prize, which was basically the Turing test. And it awarded a prize for the chat bot best able to fool a human they were human.
If you read the conversations, even the ones from 2006/2007, you'll be amazed how good the responses were. Why? Because they were basically doing the same thing as GPT3 - based on the previous five words, what is the most likely sixth word?
That's not a hard problem to solve. Given billions of gigabytes of searchable text, people can easily create systems that give plausible responses. Because all they are really doing is trying to work out what the next word (or sentence) is likely to be.
But even though this is generally considered to be the best (most Turing winning) bot out there, it's still very little better than the bots from 2006/7.
Because the next stage is really hard. GPT3 is not generalised intelligence. Deep Mind is not generalised intelligence. They are really exciting, to be sure, but they don't do what you think they do.
I know what they fucking do. They autocomplete. That's it
But you seem to think that constitutes intelligence.
It doesn't.
Who the fuck knows what truly constitutes intelligence, let alone "self awareness". It is a mystery that has perplexed philosophers for thousands of years, I have a feeling Texas car park software salesman Bobby Smithson has not personally cracked the conundrum which has confounded the greatest human minds since Aristotle
That's the point I've been making for years: 'intelligence' is exceptionally hard to define, and therefore an artificial intelligence is incredibly hard to define. I've heard experts from Cambridge argue over the definition of intelligence, and they're probably still at it.
However, it is possible to say what *isn't* intelligence. A machine putting a square peg into holes in a box at random until it fits in one isn't intelligence. And the ML stuff that you're raving about isn't intelligence either.
Head::Desk
Which is why, of course, Alan Turing invented his test, to sidestep the impossibly complex and contentious arguments around intelligence, consciousness, Theory of Mind, and all that.
Instead of trying to define intelligence and locate consciousness, you admit that it's indefinable "but you know it when you see it". Like pornography
The Turing Test is the "know it when you see it" approach: to intelligence. If the computer can convince humans it is human and intelligent, it is intelligent. This is the whole fucking point of it, tho that seems to have escaped quite a few PB-ers, probably because they are as about as smart as a ZX Spectrum
The philosophy of the Turing Test is fascinating.
However you basically don't seem to be aware of it at all.
If you have a tightly defined specification people can optimise the program they write to meet the specification that otherwise would not in any way be seen to be intelligent.
This is where the Chinese Room wankers actually have a point.
Imagine a computer program that is simply a list of all plausible things a human might say in a general conversation that is attempting to determine if the agent is human or not. For every entry in that list there is an output sentence to be repeated in response.
Boom. The list is unimaginably huge, completely rote and obviously in no way intelligent. Yet would pass the vast majority of converstations as human,
The Turing r the question with which we are concerned in this paper."
Because the example is not helpful. Of course such a machine would not be "intelligent", inasmuch as we understand the term, but that is priced in to your description
Now let me give a counter example. Imagine if I could build a complete human body from scratch with water, salt, iron, minerals, calcium, yada yada, then I put a soft squidgy computer in its skull, made out of blood and fat and grey stuff, and this brain-like computer enabled it to talk and act exactly like a human, with the same emotions and desires: totally indistinguishable
Is that human-like entity "intelligent"? We would surely treat it as such, wouldn't we? Could anyone kill it? So it has become intelligent in our eyes
And yet I have just made it out of water, salt, iron, etc, so it can't be intelligent. Can it? Yet all I have done is skipped 3 billion years of evolution, or accelerated them into a day, so maybe it IS intelligent
I see no reason why silicon and steel cannot do what iron and calcium do. Embody intelligence, or something that looks so exactly like it, the debate becomes otiose
Think he hasn't really had a proper look at them either. "Young people", he says. Most of them seem to me to be middle-aged middle-class pillocks.
The one I know about a few miles from me is a retiree millionaire in a horrible-to-heat stone farmhouse, who rents out not-very-green holiday barn conversions for up to £2.5k a week.
Prince of Wales has said world leaders gathering at the Cop26 summit should take ambitious action on climate change rather than “just talk”, and take notice of how “despairing” many young people are about the issue.
Charles said he understood why climate campaign groups such as Extinction Rebellion stage protests and block roads, but suggested they should take a less disruptive approach.
“I totally understand the frustration,” he said in an interview with the BBC.
“But it isn’t helpful, I don’t think, to do it in a way that alienates people … The difficulty is, how do you direct that frustration in a way that is more constructive rather than destructive?”
He added: “The point is, people should really notice how despairing so many young are.”
Like crooked company directors, the only thing that will make them take notice is time in a prison cell.
Maybe orange suited chain gangs cleaning up the countryside?
Almost entirely off topic my son has his first Oxford Union debate this week, that this house has no confidence in the government. A certain JRM is appearing for the government as is Geoffrey Cox QC. Should be fun.
The only time I went near the Union was to flour bomb the then President. She later became a Tory MP.
Almost entirely off topic my son has his first Oxford Union debate this week, that this house has no confidence in the government. A certain JRM is appearing for the government as is Geoffrey Cox QC. Should be fun.
I remember JRM from his first appearances at the Union. After his first few interventions an audible groan was often heard as he got up to intervene. Being a naive grammar school boy from oop nooorth I had never heard such an accent (which was even more posh then, if anything), but at least he was memorable. I got the impression that he was much better at coming up with an amusing quip than forming a long argument.
Once upon a time he said that every time he opened his mouth he lost votes for the Tories. I do wonder sometimes if that is still the case.
Think he hasn't really had a proper look at them either. "Young people", he says. Most of them seem to me to be middle-aged middle-class pillocks.
The one I know about a few miles from me is a retiree millionaire in a horrible-to-heat stone farmhouse, who rents out not-very-green holiday barn conversions for up to £2.5k a week.
Prince of Wales has said world leaders gathering at the Cop26 summit should take ambitious action on climate change rather than “just talk”, and take notice of how “despairing” many young people are about the issue.
Charles said he understood why climate campaign groups such as Extinction Rebellion stage protests and block roads, but suggested they should take a less disruptive approach.
“I totally understand the frustration,” he said in an interview with the BBC.
“But it isn’t helpful, I don’t think, to do it in a way that alienates people … The difficulty is, how do you direct that frustration in a way that is more constructive rather than destructive?”
He added: “The point is, people should really notice how despairing so many young are.”
Like crooked company directors, the only thing that will make them take notice is time in a prison cell.
Maybe orange suited chain gangs cleaning up the countryside?
I'd go for it. Preferably in North Korea or the Arctic. But I'm radicalised on Insulate Britain, having had at least one experience where promptness in getting to hospital was a life saver. Went into respiratory arrest when they made me blow into the thing to test how strong my breathing was
12 months behind prison cells, suspended for 12 months in a category C doing unpaid work in the day fitting insulation in the programme they claim does not exist.
No time to die - great movie. Second best of the Craig run after Casino Royale. Really enjoyed it, my wife did as well and she's not a huge fan of the series. Lea Seydoux looks and talks uncannily like my wife which was a bit jarring for me personally.
I'd rate it as 5th best Bond movie.
1. GoldenEye 2. Casino Royale 3. Thunderball 4. Live and Let Die 5. No Time to Die
Would be my list.
Wait.
You have Live and Let Die on there, but no Goldfinger?
That's probably within the next few tbf, I really enjoyed Live and Let Die. I know a lot of people didn't.
I really enjoyed it as a kid, but now think it's among the weaker Bond offerings (although I like the theme tune).
I would have
1. Casino Royale 2. Goldfinger 3. Thunderball 4. Goldeneye 5. The Living Daylights (which I think is a much underrated movie)
I'm hoping that NTTD gets into the top five.
My blessed Auntie Iris (RIP), the sweetest lady possible, loudly complained about Live and Let Die, to the entire Odeon (in the early 70s), "there are too many darkies"
Nick Macpherson @nickmacpherson2 · 54m The inexorable rise in bond yields is a reminder that it is as much the markets as central banks which determine the interest rates people pay. Debt interest is likely to be the fastest growing spending programme in the forthcoming spending review. #soundmoney
It will certainly be interesting to see how this pans out, rate rises expected to start this year now. From the FT:
Investors expect the Bank of England to increase rates in December for the first time since the Covid-19 crisis after policymakers sounded the alarm over inflation, pushing UK government bond yields to their highest level in two-and-a-half years.
Traders had already been tweaking bets on the BoE’s lift-off from the record low of 0.1 per cent since the central bank last month signalled concern over high inflation and said it could raise rates even before its bond-buying programme ran out at the end of the year.
Markets are pricing a rise to 0.25 per cent in the base rate by December, with a further increase to 0.5 per cent by March next year.
A combination of complacency, lethargy and Woke crap - ‘omg GPT3 might be racist’ - means the west has handed the race to AI to China, and it may already be too late to catch up.
If China dominates AI it dominates the world like no power before it
You should have learnt by now that most people aren’t interested in the really big stories.
We’ve been directly told in the last year that there is definitely ultra tech in our skies and oceans, which either belongs to adversaries of the West or non human intelligence / life forms. And everyone shrugged.
People aren’t going to listen too hard to a senior Pentagon official if he says the US has surrendered technological dominance to China and that a point will be reached (or may already have been reached) when their lead will be insurmountable. Forever.
Cognitive dissonance innit. Much more comfortable to talk about IDS’s majority instead.
Some people are INCREDIBLY resistant to the idea of artificial general intelligence. I have an extremely smart brother who is always open to new ideas but he just won't accept that this - machine intelligence - can ever happen, let alone that it is actually happening right now
Existentially, it frightens him
Or alternatively: your extremely smart brother is actually smarter than you, and realises that everything you rave about is actually fairly sh*t smoke 'n mirrors. and nowhere near 'intelligence' (*). You are staring open-mouthed, dribbling in amazement, as someone performs the three-cup trick.
I remember an august and much-missed member of this board saying that lorry drivers would not be needed due to autonomous driving. That must have been seven or eight years ago now, and we're nowhere near. In fact, we're now suffering from a shortage of drivers. Can you recall him?
The big problem with AI is the money being swilled into the trough. Billions are being spewed at it, and they need to show results. Hence smoke 'n mirrors. Just ask Musky baby where his coast-to-cast drive in a Tesla is - promised five years ago for four years ago. Yet his current tech cannot even detect emergency vehicles...
Machine learning has many uses. But they are limited in scope, and nowhere near a general intelligence. IMO that will require a massive breakthrough in tech, not the stuff we're doing atm.
You'd be amazed at an Eliza produced by drunken first-year undergrads at the West of Scotland Uni...
(*) However you define that.
Yawwwwwwn
See below. In the next ten years we will create a computer that sails through the Turing Test, at that point everything you say here will be rendered irrelevant, whether it is true or not. That *will* be AI, because we will not be able to distinguish it from human intelligence (except that it might be much cleverer and faster)
That's why the Turing Test is such a stroke of genius. It sidesteps all the philosophy and gives you a practical threshold
I don't think that this will happen. The Turing test isn't that interesting anyway. It's when a computer claims itself to be sentient that the fun starts.
This is amazing. Arguably, this is a computer passing a *kind* of Turing Test. This is Google Duplex booking a haircut, the lady at the salon believes she is talking to a human, the appointment is made, the call ends. Turing Test "passed".
But is it really passed? This guy thinks "no, not really" and I agree with him. However, I also agree with his conclusion:
"There is no way possible that we will not have a general conversational AI in the next 10 years that can speak to any human in any language about every possible topic."
With all due respect, conversation AI is not that hard, it's just giving plausible responses to inputs.
But try explaining set theory to a computer, and then get it to give you examples back.
The difference between *learning plausible responses from a giant corpus* and *learning concepts from conversation* is as wide as the Atlantc.
Now, will we get there? Of course we will. Will it happen in the next ten years? Highly, highly unlikely. Indeed, even assuming exponential intelligence growth, it may well be fifty years away.
But that is how you pass the Turing Test. The computer can only pass the tests given to it. You're shifting the goalposts
And you're also missing the point. If a computer can persuade us it is intelligent - even human - in its interactions with us, then it is, to all intents and purposes, intelligent. That's the point of the Turing Test. Is it ACTUALLY intelligent? Conscious? Thinking? Probably not, but who knows, and Immanuel Kant would struggle to give a definitive answer. And we can no more delve into its wires to seek the answer than we can open up a human brain and locate the place of consciousness
And of course once AI reaches the stage of being indistinguishable from "true" intelligence the answer to all this is, in some senses, immaterial. There will be beings walking or talking amongst us and we will not know if they are human, or not
This is going to have momentous effects, especially online
Up until last year there was a wonderful thing called the Loebner Prize, which was basically the Turing test. And it awarded a prize for the chat bot best able to fool a human they were human.
If you read the conversations, even the ones from 2006/2007, you'll be amazed how good the responses were. Why? Because they were basically doing the same thing as GPT3 - based on the previous five words, what is the most likely sixth word?
That's not a hard problem to solve. Given billions of gigabytes of searchable text, people can easily create systems that give plausible responses. Because all they are really doing is trying to work out what the next word (or sentence) is likely to be.
But even though this is generally considered to be the best (most Turing winning) bot out there, it's still very little better than the bots from 2006/7.
Because the next stage is really hard. GPT3 is not generalised intelligence. Deep Mind is not generalised intelligence. They are really exciting, to be sure, but they don't do what you think they do.
I know what they fucking do. They autocomplete. That's it
But you seem to think that constitutes intelligence.
It doesn't.
Who the fuck knows what truly constitutes intelligence, let alone "self awareness". It is a mystery that has perplexed philosophers for thousands of years, I have a feeling Texas car park software salesman Bobby Smithson has not personally cracked the conundrum which has confounded the greatest human minds since Aristotle
That's the point I've been making for years: 'intelligence' is exceptionally hard to define, and therefore an artificial intelligence is incredibly hard to define. I've heard experts from Cambridge argue over the definition of intelligence, and they're probably still at it.
However, it is possible to say what *isn't* intelligence. A machine putting a square peg into holes in a box at random until it fits in one isn't intelligence. And the ML stuff that you're raving about isn't intelligence either.
Head::Desk
Which is why, of course, Alan Turing invented his test, to sidestep the impossibly complex and contentious arguments around intelligence, consciousness, Theory of Mind, and all that.
Instead of trying to define intelligence and locate consciousness, you admit that it's indefinable "but you know it when you see it". Like pornography
The Turing Test is the "know it when you see it" approach: to intelligence. If the computer can convince humans it is human and intelligent, it is intelligent. This is the whole fucking point of it, tho that seems to have escaped quite a few PB-ers, probably because they are as about as smart as a ZX Spectrum
I disagree. The test sidesteps a central issue of whether the machine knows anything or whether it is a clever way of reducing the contents of libraries to smaller and more versatile format.
The Bodleian library's contents passes, mutatis mutandis, the Turing test. But when I go to it and read the The Critique of Pure Reason, the machine, which is the Bodleian's total contents, is merely the conveyor belt of the thoughts of Kant, not the machine. Is AI radically different in kind? or just quicker, smaller and more convenient and an infinite source of quotations.
The library does not fool into thinking the library is intelligent, nor the book. The machine does. It is a deception.
No time to die - great movie. Second best of the Craig run after Casino Royale. Really enjoyed it, my wife did as well and she's not a huge fan of the series. Lea Seydoux looks and talks uncannily like my wife which was a bit jarring for me personally.
I'd rate it as 5th best Bond movie.
1. GoldenEye 2. Casino Royale 3. Thunderball 4. Live and Let Die 5. No Time to Die
Would be my list.
I think No Time To Die started well but steadily deteriorated as the film went on - and it did go on.
I won't go into spoilers on here but though it had potential there are several things I really didn't like about it.
Think he hasn't really had a proper look at them either. "Young people", he says. Most of them seem to me to be middle-aged middle-class pillocks.
The one I know about a few miles from me is a retiree millionaire in a horrible-to-heat stone farmhouse, who rents out not-very-green holiday barn conversions for up to £2.5k a week.
Prince of Wales has said world leaders gathering at the Cop26 summit should take ambitious action on climate change rather than “just talk”, and take notice of how “despairing” many young people are about the issue.
Charles said he understood why climate campaign groups such as Extinction Rebellion stage protests and block roads, but suggested they should take a less disruptive approach.
“I totally understand the frustration,” he said in an interview with the BBC.
“But it isn’t helpful, I don’t think, to do it in a way that alienates people … The difficulty is, how do you direct that frustration in a way that is more constructive rather than destructive?”
He added: “The point is, people should really notice how despairing so many young are.”
Like crooked company directors, the only thing that will make them take notice is time in a prison cell.
Wow. I'm a bit of a closet Republican myself, but I'm not sure I'd actually have the Royal family arrested and imprisoned. (With the possible exception of Harry, for being an enormous arse.)
I appreciate neither are well regarded but I absolutely love Moonraker and Spy Who Loved Me and gladly watch them over most other Bond Movies.
The Moore Bond movies are good fun. A guilty pleasure - like admitting you like a James Blunt album.
The Spy Who Loved Me has a very comforting Britishness to it.
Other than Live and Let Die I struggle with his other movies. It was just such a huge tonal shift from serious British spy movie to sometimes slapstick British almost comedy movie, especially the latter ones where Roger Moore is ancient and very clearly out of shape. Brosnan era suffered from that too, GoldenEye was brilliant but then it quickly descended into farce with his other movies. The Craig era movies are definitely overall the second best after Sean Connery. It's a shame that we've had to wait so long for NTTD, but I'm really glad they didn't stick it onto Netflix. It deserves a big screen watch.
"Tory MP James Gray is dropped from role at St John Ambulance 'after saying "They all look the same to me" about Asian ministers Nadhim Zahawi and Sajid Javid' at reception
Tory MP for North Wiltshire James Gray made remark at charity bash Mr Gray, 66, is St John Ambulance commander and was at Parliament reception It was held to recognise 'extraordinary efforts' of volunteers and front-line staff He introduced Mr Zahawi to the stage as the Health Minister but was corrected Mr Gray then was said to comment 'They all look the same to me' to audience He acknowledged it was not appropriate and Mr Zahawi spoke to him privately Last night St John Ambulance said it did not tolerate racism in any form It has now asked him to stand down from all charity activities at once Mr Gray has denied saying remark and insists it was identity mix-up not racism"
It's entirely in character - he created a story just a few weeks ago with a joke about a bomb being sent to another MP, and seems to crop up every few years that way. He doesn't even appear to have the pseudo notoriety of some other perennial backbenchers.
He does, I remember some very senior Tory MPs from a few years ago calling James Gray an absolute copper bottomed shit (and much worse) because of this.
Almost entirely off topic my son has his first Oxford Union debate this week, that this house has no confidence in the government. A certain JRM is appearing for the government as is Geoffrey Cox QC. Should be fun.
I remember JRM from his first appearances at the Union. After his first few interventions an audible groan was often heard as he got up to intervene. Being a naive grammar school boy from oop nooorth I had never heard such an accent (which was even more posh then, if anything), but at least he was memorable. I got the impression that he was much better at coming up with an amusing quip than forming a long argument.
Once upon a time he said that every time he opened his mouth he lost votes for the Tories. I do wonder sometimes if that is still the case.
Still, it will definitely be fun.
I think my son regards JRM as something of a peacock. Interesting and curious to look at but ultimately bizarre. It will be interesting to see what he thinks tomorrow.
No time to die - great movie. Second best of the Craig run after Casino Royale. Really enjoyed it, my wife did as well and she's not a huge fan of the series. Lea Seydoux looks and talks uncannily like my wife which was a bit jarring for me personally.
I'd rate it as 5th best Bond movie.
1. GoldenEye 2. Casino Royale 3. Thunderball 4. Live and Let Die 5. No Time to Die
Would be my list.
Wait.
You have Live and Let Die on there, but no Goldfinger?
If I had to guess I'd say we're on course for 2-2.5C of warming at its highest point.
That won't be fun and will cause a lot of extreme weather, migration and political disruption (remember: we are already at 1.2C of warming now) and the loss of quite a few marginal species but we won't become "extinct".
Far from it.
That would probably be enough warming to trigger a complete melt of the Greenland Ice Sheet, which would have state-ending consequences for many nations - how does Egypt survive without the Nile Delta?
I'm beginning to doubt whether democracy will survive the expense, hardship and dislocations of the resulting adjustments. Food security in particular should be a major concern for the British government.
Disruption to agriculture will increase and, as we've seen with Covid, the first instinct of many countries will be to hoard supplies and ban exports.
Yes!! Anyone with the ear of government get this message across to those thickos please. Tell them to go and watch an old video of Jason Donovan in his technicoloured coat. 7 fat cows and 7 skinny cows.
It should not be necessary for citizens with foresight to consider how best to stockpile nutrition. The government should be building a national strategic nutrition reserve. I don’t care whether it’s a fully nationalised endeavour or one outsourced to the big supermarkets. And they should be considering how we might close our nutrition import gap. And they should of course be doing the same thing for energy.
It’s the national equivalent of income protection or critical illness insurance. It’s depressing how poor our strategic resilience has become.
A combination of complacency, lethargy and Woke crap - ‘omg GPT3 might be racist’ - means the west has handed the race to AI to China, and it may already be too late to catch up.
If China dominates AI it dominates the world like no power before it
You should have learnt by now that most people aren’t interested in the really big stories.
We’ve been directly told in the last year that there is definitely ultra tech in our skies and oceans, which either belongs to adversaries of the West or non human intelligence / life forms. And everyone shrugged.
People aren’t going to listen too hard to a senior Pentagon official if he says the US has surrendered technological dominance to China and that a point will be reached (or may already have been reached) when their lead will be insurmountable. Forever.
Cognitive dissonance innit. Much more comfortable to talk about IDS’s majority instead.
Some people are INCREDIBLY resistant to the idea of artificial general intelligence. I have an extremely smart brother who is always open to new ideas but he just won't accept that this - machine intelligence - can ever happen, let alone that it is actually happening right now
Existentially, it frightens him
Or alternatively: your extremely smart brother is actually smarter than you, and realises that everything you rave about is actually fairly sh*t smoke 'n mirrors. and nowhere near 'intelligence' (*). You are staring open-mouthed, dribbling in amazement, as someone performs the three-cup trick.
I remember an august and much-missed member of this board saying that lorry drivers would not be needed due to autonomous driving. That must have been seven or eight years ago now, and we're nowhere near. In fact, we're now suffering from a shortage of drivers. Can you recall him?
The big problem with AI is the money being swilled into the trough. Billions are being spewed at it, and they need to show results. Hence smoke 'n mirrors. Just ask Musky baby where his coast-to-cast drive in a Tesla is - promised five years ago for four years ago. Yet his current tech cannot even detect emergency vehicles...
Machine learning has many uses. But they are limited in scope, and nowhere near a general intelligence. IMO that will require a massive breakthrough in tech, not the stuff we're doing atm.
You'd be amazed at an Eliza produced by drunken first-year undergrads at the West of Scotland Uni...
(*) However you define that.
Yawwwwwwn
See below. In the next ten years we will create a computer that sails through the Turing Test, at that point everything you say here will be rendered irrelevant, whether it is true or not. That *will* be AI, because we will not be able to distinguish it from human intelligence (except that it might be much cleverer and faster)
That's why the Turing Test is such a stroke of genius. It sidesteps all the philosophy and gives you a practical threshold
I don't think that this will happen. The Turing test isn't that interesting anyway. It's when a computer claims itself to be sentient that the fun starts.
This is amazing. Arguably, this is a computer passing a *kind* of Turing Test. This is Google Duplex booking a haircut, the lady at the salon believes she is talking to a human, the appointment is made, the call ends. Turing Test "passed".
But is it really passed? This guy thinks "no, not really" and I agree with him. However, I also agree with his conclusion:
"There is no way possible that we will not have a general conversational AI in the next 10 years that can speak to any human in any language about every possible topic."
With all due respect, conversation AI is not that hard, it's just giving plausible responses to inputs.
But try explaining set theory to a computer, and then get it to give you examples back.
The difference between *learning plausible responses from a giant corpus* and *learning concepts from conversation* is as wide as the Atlantc.
Now, will we get there? Of course we will. Will it happen in the next ten years? Highly, highly unlikely. Indeed, even assuming exponential intelligence growth, it may well be fifty years away.
But that is how you pass the Turing Test. The computer can only pass the tests given to it. You're shifting the goalposts
And you're also missing the point. If a computer can persuade us it is intelligent - even human - in its interactions with us, then it is, to all intents and purposes, intelligent. That's the point of the Turing Test. Is it ACTUALLY intelligent? Conscious? Thinking? Probably not, but who knows, and Immanuel Kant would struggle to give a definitive answer. And we can no more delve into its wires to seek the answer than we can open up a human brain and locate the place of consciousness
And of course once AI reaches the stage of being indistinguishable from "true" intelligence the answer to all this is, in some senses, immaterial. There will be beings walking or talking amongst us and we will not know if they are human, or not
This is going to have momentous effects, especially online
Up until last year there was a wonderful thing called the Loebner Prize, which was basically the Turing test. And it awarded a prize for the chat bot best able to fool a human they were human.
If you read the conversations, even the ones from 2006/2007, you'll be amazed how good the responses were. Why? Because they were basically doing the same thing as GPT3 - based on the previous five words, what is the most likely sixth word?
That's not a hard problem to solve. Given billions of gigabytes of searchable text, people can easily create systems that give plausible responses. Because all they are really doing is trying to work out what the next word (or sentence) is likely to be.
But even though this is generally considered to be the best (most Turing winning) bot out there, it's still very little better than the bots from 2006/7.
Because the next stage is really hard. GPT3 is not generalised intelligence. Deep Mind is not generalised intelligence. They are really exciting, to be sure, but they don't do what you think they do.
I know what they fucking do. They autocomplete. That's it
But you seem to think that constitutes intelligence.
It doesn't.
Who the fuck knows what truly constitutes intelligence, let alone "self awareness". It is a mystery that has perplexed philosophers for thousands of years, I have a feeling Texas car park software salesman Bobby Smithson has not personally cracked the conundrum which has confounded the greatest human minds since Aristotle
That's the point I've been making for years: 'intelligence' is exceptionally hard to define, and therefore an artificial intelligence is incredibly hard to define. I've heard experts from Cambridge argue over the definition of intelligence, and they're probably still at it.
However, it is possible to say what *isn't* intelligence. A machine putting a square peg into holes in a box at random until it fits in one isn't intelligence. And the ML stuff that you're raving about isn't intelligence either.
Head::Desk
Which is why, of course, Alan Turing invented his test, to sidestep the impossibly complex and contentious arguments around intelligence, consciousness, Theory of Mind, and all that.
Instead of trying to define intelligence and locate consciousness, you admit that it's indefinable "but you know it when you see it". Like pornography
The Turing Test is the "know it when you see it" approach: to intelligence. If the computer can convince humans it is human and intelligent, it is intelligent. This is the whole fucking point of it, tho that seems to have escaped quite a few PB-ers, probably because they are as about as smart as a ZX Spectrum
I disagree. The test sidesteps a central issue of whether the machine knows anything or whether it is a clever way of reducing the contents of libraries to smaller and more versatile format.
The Bodleian library's contents passes, mutatis mutandis, the Turing test. But when I go to it and read the The Critique of Pure Reason, the machine, which is the Bodleian's total contents, is merely the conveyor belt of the thoughts of Kant, not the machine. Is AI radically different in kind? or just quicker, smaller and more convenient and an infinite source of quotations.
Indeed. But Turing devised the test explicitly and knowingly to "side-step" these eternal philosophical debates. It's in his original paper, which I quoted below
"I do not wish to give the impression that I think there is no mystery about consciousness. There is, for instance, something of a paradox connected with any attempt to localise it. But I do not think these mysteries necessarily need to be solved before we can answer the question with which we are concerned in this paper"
Basically: we're never going to get an answer to What is intelligence, or What is consciousness, but that should not stop us assessing whether we have arrived at machine intelligence, so here's a practical way of doing it
You can criticise the Turing Test for being simplistic, or a dead end, or outdated, or plain wrong-headed, but at least he had a go and no one has come up with an obviously better approach, as far as I know
No time to die - great movie. Second best of the Craig run after Casino Royale. Really enjoyed it, my wife did as well and she's not a huge fan of the series. Lea Seydoux looks and talks uncannily like my wife which was a bit jarring for me personally.
I'd rate it as 5th best Bond movie.
1. GoldenEye 2. Casino Royale 3. Thunderball 4. Live and Let Die 5. No Time to Die
Would be my list.
Wait.
You have Live and Let Die on there, but no Goldfinger?
And where's On Her Majesty's Secret Service?!
That's a really good Bond - but top five? Tricky.
I haven't seen NTTD, but of the rest, I'd argue it's the best bond film (with Casino Royale a close 2nd).
No time to die - great movie. Second best of the Craig run after Casino Royale. Really enjoyed it, my wife did as well and she's not a huge fan of the series. Lea Seydoux looks and talks uncannily like my wife which was a bit jarring for me personally.
I'd rate it as 5th best Bond movie.
1. GoldenEye 2. Casino Royale 3. Thunderball 4. Live and Let Die 5. No Time to Die
Would be my list.
I think No Time To Die started well but steadily deteriorated as the film went on - and it did go on.
I won't go into spoilers on here but though it had potential there are several things I really didn't like about it.
Yeah the length was the one major drawback. They could have edited down a fair chunk of it, especially the bits in London. Despite the length the baddie felt seriously undercooked and underused as well.
Comments
Or maybe you would...
Robert Reich
@RBReich
“The situation in America is such that I would favor states seceding from the union to form their own separate country.”
Trump voters: 52 percent agree, 25 percent strongly.
Biden voters: 41 percent agree, 18 percent strongly.
===
Erm, isn't it the left of Trump neo-fascism states that will be leaving???
"WuDao 2.0 Generates The First Virtual Student"
https://analyticsdrift.com/wudao-2-0-generates-the-first-virtual-student/
Remember too every state of the old Confederacy voted for Trump in 2020 except for Virginia and Georgia (and Georgia voted for Trump in 2016)
However, it is possible to say what *isn't* intelligence. A machine putting a square peg into holes in a box at random until it fits in one isn't intelligence. And the ML stuff that you're raving about isn't intelligence either.
https://www.swindonadvertiser.co.uk/news/922696.mp-had-affair-while-wife-battled-cancer/
https://www.thetimes.co.uk/article/tory-mp-had-affair-while-his-wife-suffered-cancer-qsdqr3kldm8
He's quite unpleasant.
Which is why, of course, Alan Turing invented his test, to sidestep the impossibly complex and contentious arguments around intelligence, consciousness, Theory of Mind, and all that.
Instead of trying to define intelligence and locate consciousness, you admit that it's indefinable "but you know it when you see it". Like pornography
The Turing Test is the "know it when you see it" approach: to intelligence. If the computer can convince humans it is human and intelligent, it is intelligent. This is the whole fucking point of it, tho that seems to have escaped quite a few PB-ers, probably because they are as about as smart as a ZX Spectrum
That said, his seat is proposed to be chopped up and joined with parts of Gloucestershire.
I'm also wary when people talk about the size of the datasets going into it. Large numbers sound impressive. Datasets are useful for ML; less so for real AI. And GIGO is a massive factor with datasets. Where are the datasets coming from? Have they been verified? Are they fit for this purpose?
Also, the claims are from a researcher, and there is zero independent verification of the work.
Until we get such, it is all just smoke 'n mirrors.
Also, Turing did not envisage the amount of data that could be used to create a system that could fake intelligence.
Why does this constant over-hyping of AI matter? Because pretending dumb systems are intelligent can have massive real-world effects:
https://www.technologyreview.com/2019/01/21/137783/algorithms-criminal-justice-ai/
Prince Charles is on the side of insulate Britain.
https://www.gbnews.uk/news/prince-charles-understands-extinction-rebellion-and-insulate-britain-protests/139455
Can't wait for him to become Monarch.
The word "all" is the give-away, indicating a racist joke; there were only two of them.
If he'd simply said "They (both) look the same to me", an innocent (albeit stupid because they don't look alike) error is plausible.
And it was only last week that he 'joked' about delivering a bomb to Anneliese Dodds' office.
However you basically don't seem to be aware of it at all.
If you have a tightly defined specification people can optimise the program they write to meet the specification that otherwise would not in any way be seen to be intelligent.
This is where the Chinese Room wankers actually have a point.
Imagine a computer program that is simply a list of all plausible things a human might say in a general conversation that is attempting to determine if the agent is human or not. For every entry in that list there is an output sentence to be repeated in response.
Boom. The list is unimaginably huge, completely rote and obviously in no way intelligent. Yet would pass the vast majority of converstations as human,
“Understands” - didn’t someone say they could understand the motives of Palestinian suicide bombers?
God intended America to be promised land for European Christians (% AGREE):
30% all
52% wh evangelical Prot
53% Republican
52% trust Fox News
67% trust far right news
65% QAnon believer
@PRRIpoll
Cc @ndrewwhitehead
@socofthesacred
@KevinMKruse
https://twitter.com/robertpjones/status/1447601670595223557
I think you're thinking of Jenny Tonge.
https://en.wikipedia.org/wiki/Jenny_Tonge,_Baroness_Tonge#Palestinian_suicide_bombers_(2004)
I'd rate it as 5th best Bond movie.
1. GoldenEye
2. Casino Royale
3. Thunderball
4. Live and Let Die
5. No Time to Die
Would be my list.
"The power and appeal of the Turing test derives from its simplicity. The philosophy of mind, psychology, and modern neuroscience have been unable to provide definitions of "intelligence" and "thinking" that are sufficiently precise and general to be applied to machines. Without such definitions, the central questions of the philosophy of artificial intelligence cannot be answered. The Turing test, even if imperfect, at least provides something that can actually be measured. As such, it is a pragmatic attempt to answer a difficult philosophical question."
Turing himself, in his original paper, proposed his test so as to avoid the impossible enigma of "consciousness"
"I do not wish to give the impression that I think there is no mystery about consciousness. There is, for instance, something of a paradox connected with any attempt to localise it. But I do not think these mysteries necessarily need to be solved before we can answer the question with which we are concerned in this paper."
https://en.wikipedia.org/wiki/Turing_test
While 52% of white evangelicals agree, only 34% of white mainline Protestants and 37% of white Catholics agree
If he could vote Prince Charles would probably vote LD or Green
Surely the time must be nearing when he devotes himself full time to his true calling.
You have Live and Let Die on there, but no Goldfinger?
http://iaindale.blogspot.com/2009/05/exclusive-james-gray-photo-of-dying.html
I would have
1. Casino Royale
2. Goldfinger
3. Thunderball
4. Goldeneye
5. The Living Daylights (which I think is a much underrated movie)
I'm hoping that NTTD gets into the top five.
I shall disappear without Tracey.
Good night.
Think he hasn't really had a proper look at them either. "Young people", he says. Most of them seem to me to be middle-aged middle-class pillocks.
The one I know about a few miles from me is a retiree millionaire in a horrible-to-heat stone farmhouse, who rents out not-very-green holiday barn conversions for up to £2.5k a week.
Prince of Wales has said world leaders gathering at the Cop26 summit should take ambitious action on climate change rather than “just talk”, and take notice of how “despairing” many young people are about the issue.
Charles said he understood why climate campaign groups such as Extinction Rebellion stage protests and block roads, but suggested they should take a less disruptive approach.
“I totally understand the frustration,” he said in an interview with the BBC.
“But it isn’t helpful, I don’t think, to do it in a way that alienates people … The difficulty is, how do you direct that frustration in a way that is more constructive rather than destructive?”
He added: “The point is, people should really notice how despairing so many young are.”
Like crooked company directors, the only thing that will make them take notice is time in a prison cell.
Brooke Bond
Basildon Bond
Premium Bond
Chemical Bond
Stretcher Bond
Sorry, James Bond just doesn't make the list.
Now let me give a counter example. Imagine if I could build a complete human body from scratch with water, salt, iron, minerals, calcium, yada yada, then I put a soft squidgy computer in its skull, made out of blood and fat and grey stuff, and this brain-like computer enabled it to talk and act exactly like a human, with the same emotions and desires: totally indistinguishable
Is that human-like entity "intelligent"? We would surely treat it as such, wouldn't we? Could anyone kill it? So it has become intelligent in our eyes
And yet I have just made it out of water, salt, iron, etc, so it can't be intelligent. Can it? Yet all I have done is skipped 3 billion years of evolution, or accelerated them into a day, so maybe it IS intelligent
I see no reason why silicon and steel cannot do what iron and calcium do. Embody intelligence, or something that looks so exactly like it, the debate becomes otiose
Unfortunately, they then go to space, and it all falls apart.
Or former soccer manager, John Bond ?
The Spy Who Loved Me has a very comforting Britishness to it.
I'm really looking forward to when he inevitably refuses to give royal assent to some bill he doesn't look.
Once upon a time he said that every time he opened his mouth he lost votes for the Tories. I do wonder sometimes if that is still the case.
Still, it will definitely be fun.
12 months behind prison cells, suspended for 12 months in a category C doing unpaid work in the day fitting insulation in the programme they claim does not exist.
Then she walked out.
Autre temps, autre moeurs
The Bodleian library's contents passes, mutatis mutandis, the Turing test. But when I go to it and read the The Critique of Pure Reason, the machine, which is the Bodleian's total contents, is merely the conveyor belt of the thoughts of Kant, not the machine. Is AI radically different in kind? or just quicker, smaller and more convenient and an infinite source of quotations.
The library does not fool into thinking the library is intelligent, nor the book. The machine does. It is a deception.
I won't go into spoilers on here but though it had potential there are several things I really didn't like about it.
Bond Bug,
It should not be necessary for citizens with foresight to consider how best to stockpile nutrition. The government should be building a national strategic nutrition reserve. I don’t care whether it’s a fully nationalised endeavour or one outsourced to the big supermarkets. And they should be considering how we might close our nutrition import gap. And they should of course be doing the same thing for energy.
It’s the national equivalent of income protection or critical illness insurance. It’s depressing how poor our strategic resilience has become.
https://www.bbc.co.uk/iplayer/episode/p02sx893/premium-bond-with-mark-gatiss-and-matthew-sweet
"I do not wish to give the impression that I think there is no mystery about consciousness. There is, for instance, something of a paradox connected with any attempt to localise it. But I do not think these mysteries necessarily need to be solved before we can answer the question with which we are concerned in this paper"
Basically: we're never going to get an answer to What is intelligence, or What is consciousness, but that should not stop us assessing whether we have arrived at machine intelligence, so here's a practical way of doing it
You can criticise the Turing Test for being simplistic, or a dead end, or outdated, or plain wrong-headed, but at least he had a go and no one has come up with an obviously better approach, as far as I know