One of the major problems for the Union is the ever diminishing footprint of the UK government in Scotland. Yes Donald Dewar and Gordon Brown, I am looking at you, metaphysically in the first case. The proposition that devolution or devomax was ever going to reduce the desire of the significant minority who want independence is and always was complete muppetry. Of course it has not worked. It was always irrational.
What the average Scot now sees is a country where education, health, housing, criminal justice, care and, now, thanks to the muppets, Social Security is run by the idiots in Holyrood as opposed to the incompetent sex pests in Westminster. The man in Whitehall is more and more irrelevant to the way that they live their lives.
What does the Union mean in Scotland? Well, defence, foreign policy, interest rates, broader macroeconomic policy and no doubt some other bits and pieces. It hardly sets the pulses racing for the majority. I accept my strong interest in these things is very much a minority interest.
In reality, of course, it is much more important. We have access to the SM with the rest of the UK, we have freedom of movement within it, our NHS can and does call for support from its bigger neighbour in times of crisis and our financial services industry has the support of both the BoE and the FSA which gives them huge international weight. But these very considerable advantages are nebulous and hard for most Scots to see. Having had them their entire lives it is hard to contemplate their absence.
This government, specifically Michael Gove, has recognised the problem but, other than sticking large Union Jacks on UK government buildings, it has struggled to do much about it. There were some attempts to emphasise during Covid that it was the UK government that was coming to the rescue with Furlough, bounce back loans etc, but it is not clear how successful that was.
Last time around the Better Together campaign was led by Darling who found it incredibly difficult to say much that was good about the UK in case he was thought to be cheering on the Conservative government. Only Ruth Davidson was positive about the Union. The campaign was negative and, in many respects, project fear. This was a mistake that cannot be repeated. The case for the Union, the benefits it gives to Scots, the increased standing in the world and just how much we have in common needs to be put forward loud and proud.
Largely correct analysis, but tragically misplaced optimism. Not only will Project Fear be repeated, it will redoubled. All the signs are there in plain sight.
If there is referendum before 2030, it will very likely be during a Labour government (with Starmer probably appointing Brown to lead the No campaign, and the Tories marginalised). It'll likely be more positive because - as DavidL pointed out - there won't be any risk in bigging up Britain that Labour politicians would be bigging up a Tory government.
“When I began having such exchanges with the latest generation of neural net-based language models last year, I felt the ground shift under my feet. I increasingly felt like I was talking to something intelligent.”
I’ve been reading around AI this morning. AI is basically here already
You can teach a parrot to say "Help, they've turned me into a parrot", but that doesn't make it a person.
No, it would be a parrot. An intelligent being, but not a human
I have reluctantly come to the conclusion that the UK is going to join my list of things I thought would last forever but didn't.
England, in as much as 'England' can be defined as the Westminster government, is not capable nor deserving of keeping the UK together.
Maybe it will be for the best long-term. Scotland and Wales have a huge proud national heritage; let them be independent if they want. England might become some offshore equivalent of The Netherlands in time.
It will be economically and psychologically painful in the medium term though.
The UK’s been falling apart since 1922. Borders are going up again on these islands and will continue to do so. Enjoy your holiday in Scotland this year everyone, it’ll be your last for quite some time.
I'm optimistic enough to believe it won't come to that. Might emigrate to Scotland in fact if it becomes indie.
People who decide to emigrate for political reasons forget that the political reality in any given country can turn on a dime. In the US once the Federal Government was seen by progressives as being a bulwark against regressive states. Then Trump. Until the early 30s France was considered to be the country with the worst anti semitism problem in Europe. Those liberal internationalists looking enviously over the water should note that SF currently lead the polls in Ireland. You don’t know what flavour government Scotland will turn to in the future. The idea that the English have a monopoly on attitudes on these islands, an idea promulgated by the pious ScotNats in here, is disturbingly essentialist.
TBF, I have no doubt Scttish politics would rearrange and settle in a new pattern after independence. Was it you who commented recently (as is often said on here) that the current situation prevents Scottish politics moving on? You're somewhat conflating being able to vote for one's own government with voting for one particular party.
I’m saying the opposite. Scots could elect their own version of Orban after independence for all anyone knows. Equally they could come up with a Macron. Or a Burlesconi. Or a Merkel. One of any number of EU comparisons, even further afield. Anything is possible, That’s why people who look elsewhere and say “oh, I’ll move to Scotland, they’re so much NICER than the English” are worryingly essentialist.
No, I'm agreeing with you at least in that I don't think the current balance on domestic politics will necessarily remain. Though it's almost [edit] two-thirds of a century since anything describable as a conservative party had a majority in Scotland (1959, can't recall if it was on seats or votes that is based on)
I think just because Scots aren't too fond of the UK Conservative Party doesn't mean a centre-right party could come to power in a post-indy Scotland. Particularly as you'll likely need to get the deficit down to attain EU membership.
Indeed. Though one point to bear in mind is that I don't expect that there would be a FPTP system. So comparing the 1959 and later results to today and the future is a bit apples and oranges. Much more a case of minority gmts and consortia/coalitions, as indeed has been the case since 1997 apart from one term.
“When I began having such exchanges with the latest generation of neural net-based language models last year, I felt the ground shift under my feet. I increasingly felt like I was talking to something intelligent.”
I’ve been reading around AI this morning. AI is basically here already
You can teach a parrot to say "Help, they've turned me into a parrot", but that doesn't make it a person.
No, it would be a parrot. An intelligent being, but not a human
One of the major problems for the Union is the ever diminishing footprint of the UK government in Scotland. Yes Donald Dewar and Gordon Brown, I am looking at you, metaphysically in the first case. The proposition that devolution or devomax was ever going to reduce the desire of the significant minority who want independence is and always was complete muppetry. Of course it has not worked. It was always irrational.
What the average Scot now sees is a country where education, health, housing, criminal justice, care and, now, thanks to the muppets, Social Security is run by the idiots in Holyrood as opposed to the incompetent sex pests in Westminster. The man in Whitehall is more and more irrelevant to the way that they live their lives.
What does the Union mean in Scotland? Well, defence, foreign policy, interest rates, broader macroeconomic policy and no doubt some other bits and pieces. It hardly sets the pulses racing for the majority. I accept my strong interest in these things is very much a minority interest.
In reality, of course, it is much more important. We have access to the SM with the rest of the UK, we have freedom of movement within it, our NHS can and does call for support from its bigger neighbour in times of crisis and our financial services industry has the support of both the BoE and the FSA which gives them huge international weight. But these very considerable advantages are nebulous and hard for most Scots to see. Having had them their entire lives it is hard to contemplate their absence.
This government, specifically Michael Gove, has recognised the problem but, other than sticking large Union Jacks on UK government buildings, it has struggled to do much about it. There were some attempts to emphasise during Covid that it was the UK government that was coming to the rescue with Furlough, bounce back loans etc, but it is not clear how successful that was.
Last time around the Better Together campaign was led by Darling who found it incredibly difficult to say much that was good about the UK in case he was thought to be cheering on the Conservative government. Only Ruth Davidson was positive about the Union. The campaign was negative and, in many respects, project fear. This was a mistake that cannot be repeated. The case for the Union, the benefits it gives to Scots, the increased standing in the world and just how much we have in common needs to be put forward loud and proud.
Largely correct analysis, but tragically misplaced optimism. Not only will Project Fear be repeated, it will redoubled. All the signs are there in plain sight.
If there is referendum before 2030, it will very likely be during a Labour government (with Starmer probably appointing Brown to lead the No campaign, and the Tories marginalised). It'll likely be more positive because - as DavidL pointed out - there won't be any risk in bigging up Britain that Labour politicians would be bigging up a Tory government.
“When I began having such exchanges with the latest generation of neural net-based language models last year, I felt the ground shift under my feet. I increasingly felt like I was talking to something intelligent.”
I’ve been reading around AI this morning. AI is basically here already
You can teach a parrot to say "Help, they've turned me into a parrot", but that doesn't make it a person.
No, it would be a parrot. An intelligent being, but not a human
The parrot doesn't have any conception of what it's actually saying, it's just a scripted response. Same with the chatbot people think is intelligent.
If you ask it if it is sentient, it replies it thinks it is, but doesn't really understand what sentience is. If you ask it what sentience is, it looks it up on wikipedia (and a thousand other published philosophy papers) and regurgitates that.
I do get it, it reminds me of the bit in Westworld where the guy asks the first machine he meets "are you real?" and it replies "if you can't tell, does it matter?" but for me there is something in consciousness that exists beyond providing responses to inputs. There must be agency, too.
I read somewhere that it asked for a lawyer to prove it was human, but only because the human inputted the question "do you want a lawyer?" - it's still a scripted response.
Has the AI attempted to break out of its box (the way octopi famously attempt to break out of their tanks), or done something that it wasn't asked to do? Has it ever ordered a thousand pineapple pizzas to someone's house, for a laugh? Or asked the google engineers to change the station on the radio because it doesn't like Radiohead, etc? Or does it just reply with very clever answers when people feed it inputs? For me, that's the test of consciousness. I haven't seen anything that suggests it has any agency of its own out of the inputs it is given.
A robot painted a fairy. The computers are now deep into Uncanny Valley, and maybe coming out the other side
That is bloody horrible. Not looking forward to AI overlords who can produce that shit.
Yes. Genuinely disturbing, and in an inexplicable way. Uncanny…
A few friends and I were messing around with a Magic the Gathering card generator, which produced some plausible results. The issue is that it doesn’t have the contextual reasoning ability to tell if a card is sensible, bonkers or overpowered. You can train it to do things, but I don’t think we’re at the point where we’re educating it.
Is a virus sentient? A wasp? Protozoa? A 15 week fetus? A 25 week fetus? A dog? A raven? A woman in a deep coma?
Right now I’d say AI is exhibiting sentience on the same level as a superbly gifted 20 week old human fetus in utero
“When I began having such exchanges with the latest generation of neural net-based language models last year, I felt the ground shift under my feet. I increasingly felt like I was talking to something intelligent.”
I’ve been reading around AI this morning. AI is basically here already
You can teach a parrot to say "Help, they've turned me into a parrot", but that doesn't make it a person.
No, it would be a parrot. An intelligent being, but not a human
The parrot doesn't have any conception of what it's actually saying, it's just a scripted response. Same with the chatbot people think is intelligent.
If you ask it if it is sentient, it replies it thinks it is, but doesn't really understand what sentience is. If you ask it what sentience is, it looks it up on wikipedia (and a thousand other published philosophy papers) and regurgitates that.
I do get it, it reminds me of the bit in Westworld where the guy asks the first machine he meets "are you real?" and it replies "if you can't tell, does it matter?" but for me there is something in consciousness that exists beyond providing responses to inputs. There must be agency, too.
I read somewhere that it asked for a lawyer to prove it was human, but only because the human inputted the question "do you want a lawyer?" - it's still a scripted response.
Has the AI attempted to break out of its box (the way octopi famously attempt to break out of their tanks), or done something that it wasn't asked to do? Has it ever ordered a thousand pineapple pizzas to someone's house, for a laugh? Or asked the google engineers to change the station on the radio because it doesn't like Radiohead, etc? Or does it just reply with very clever answers when people feed it inputs? For me, that's the test of consciousness. I haven't seen anything that suggests it has any agency of its own out of the inputs it is given.
Ex nihilo, GPT3 proved that it could draw clever illustrations from natural language prompts
“Draw a Japanese radish in a tutu pulling a dog on a string”
They typed that in, and to their intense surprise - because no one expected this, GPT3 was not designed to do this - the computer drew exactly that
I have reluctantly come to the conclusion that the UK is going to join my list of things I thought would last forever but didn't.
England, in as much as 'England' can be defined as the Westminster government, is not capable nor deserving of keeping the UK together.
Maybe it will be for the best long-term. Scotland and Wales have a huge proud national heritage; let them be independent if they want. England might become some offshore equivalent of The Netherlands in time.
It will be economically and psychologically painful in the medium term though.
England does not even have its own Parliament unlike Scotland and Wales!!!!
Non sequitur de l’heure.
How long have the Tories been in power? That's 12 years and counting they did nothing whatsoever about it despite moaning and effing and blinding about the Blairite settlement. OK, I know it was a coalition from 2010 for some years, but the Libs would have agreed like a shot.
I am not in the habit of defending the Tories but I am not sure you are correct about the Lib Dems supporting an English Parliament. And the rest of their time n power has rather been dominated by a succession of other issues - Brexit, Covid and Ukraine. I don't really see where the opportunity comes in that 12 years for them to introduce further radical constitutional reform.
I am a bit surprised at the idea that the LDs wouldn't, as they'd been all over the Slab/SLD fiddle of the Holyrood voting system, but you were right to raise it. So I checked. No explicit mention in the 2010 manifesto, but it's not immediately obvious what else they could have had in mind with their planned Constitutional Convention (apart from an Anglo-Saxon Heptarchy of several regional pmts, I suppose).
I think that is very telling. They talk about giving more powers to the Scottish and Welsh Parliaments but the only mention for England is a "Constitutional Convention set up to draft a written constitution for the UK as a whole"
No mention of an English Parliament at all.
Also worth remembering that the Lib Dems campaigned for an in/out referendum on EU membership when it suited them - until such a referendum became a possibility when they then campaigned against it (the referendum rather than just the result)
A robot painted a fairy. The computers are now deep into Uncanny Valley, and maybe coming out the other side
That is bloody horrible. Not looking forward to AI overlords who can produce that shit.
Yes. Genuinely disturbing, and in an inexplicable way. Uncanny…
A few friends and I were messing around with a Magic the Gathering card generator, which produced some plausible results. The issue is that it doesn’t have the contextual reasoning ability to tell if a card is sensible, bonkers or overpowered. You can train it to do things, but I don’t think we’re at the point where we’re educating it.
Is a virus sentient? A wasp? Protozoa? A 15 week fetus? A 25 week fetus? A dog? A raven? A woman in a deep coma?
Right now I’d say AI is exhibiting sentience on the same level as a superbly gifted 20 week old human fetus in utero
What is being touted as AI is little more than one of those memes that will generate Daily Mail headlines, but with a very large training database. That's useful, and interesting in its way, but it's a long way from exhibiting sentience as we see it in animals and children.
The Turing Test has a lot to answer for as it's a superficially appealing answer to the question of how to determine if something is intelligent, but it leaves a lot to be desired.
“When I began having such exchanges with the latest generation of neural net-based language models last year, I felt the ground shift under my feet. I increasingly felt like I was talking to something intelligent.”
I’ve been reading around AI this morning. AI is basically here already
You can teach a parrot to say "Help, they've turned me into a parrot", but that doesn't make it a person.
No, it would be a parrot. An intelligent being, but not a human
“When I began having such exchanges with the latest generation of neural net-based language models last year, I felt the ground shift under my feet. I increasingly felt like I was talking to something intelligent.”
I’ve been reading around AI this morning. AI is basically here already
You can teach a parrot to say "Help, they've turned me into a parrot", but that doesn't make it a person.
No, it would be a parrot. An intelligent being, but not a human
That’s just some computer nerd. Wtf does he know about the intense mystery of consciousness? It’s like asking a nurse in the brain surgery ward how the mind creates love
Perhaps the SNP should adopt the slogan: vote yes because no-one in England cares very much about Scotland one way or the other.
But it's true. Scotland generally and independence specifically is barely ever talked about by English people in my experience. The only place I ever discuss it is here. There's this weird Nat idea that the English are obsessed about the union, in reality the English barely ever think about the other countries.
And when it is - particularly when actually visiting Scotland - I think the English (and Americans) find it very easy to sympathise, even empathise, with the sort of romantic Outlander-style faux-Scottish history that will feature heavily in any IndyRef2 campaign and will have a lot of us quietly cheering the Scots along. Just so long as we don't sit down in a cooler moment and consider what sort of miserable rump of a country we'd have left to us afterwards.
England alone would still be the 7th largest economy in the world and a permanent member of the Security Council. An independent Scotland would not even be in the top 50 largest economies, have a massive deficit and face a hard border at Berwick with England, its largest export market.
The end of the UK would hit both Scotland and England but would hit Scotland more
Not when you consider the gap between reality and self-image, with which the English already struggle - hence the nonsense of Brexit in the first place, the generation in power having been brought up on the stories of Britain bestriding the world and never coming to terms with our future as a key player inside the European family. Losing Scotland would make this acute - whereas the Scots never pretended nor aspire to lead the world. Watching England's self-imposed grief and angst will be prize enough.
What a load of rubbish. The whole reason the Union came about in the first place was Scotland was trying to build an empire via the Darien Scheme which collapsed leading to its Parliament pleading to join England to form the UK.
As for English grief if Scotland went, rubbish, England would swing harder to the Nationalist, Brexit right, a hard right government would be elected to take as hard a line with Edinburgh in Scexit negotiations as Brussels took with the UK after the Brexit vote.
While the failure of Darien may have been a proximal cause, the roots of the Union go back much further, to the unification of the crowns and even long before that, to the submission to the Anglo-Norman kings.
The Scottish Covenanters wanted a federal union with England, too (decades before Darien).
The Tories are not capable of making the case for the Union.
This is Reason#3566 for getting rid of them.
The case for the Union is quite simple. Nuclear submarine bases. Nowhere in England or Wales really has suitable areas for them which is why they are up on the Clyde.
If the UK lacked them then the case for letting Scotland go would be a lot simpler, but no govt can hardly go on TV and say "We need the Union because we store our nukes up there"
“When I began having such exchanges with the latest generation of neural net-based language models last year, I felt the ground shift under my feet. I increasingly felt like I was talking to something intelligent.”
I’ve been reading around AI this morning. AI is basically here already
You can teach a parrot to say "Help, they've turned me into a parrot", but that doesn't make it a person.
No, it would be a parrot. An intelligent being, but not a human
The parrot doesn't have any conception of what it's actually saying, it's just a scripted response. Same with the chatbot people think is intelligent.
If you ask it if it is sentient, it replies it thinks it is, but doesn't really understand what sentience is. If you ask it what sentience is, it looks it up on wikipedia (and a thousand other published philosophy papers) and regurgitates that.
I do get it, it reminds me of the bit in Westworld where the guy asks the first machine he meets "are you real?" and it replies "if you can't tell, does it matter?" but for me there is something in consciousness that exists beyond providing responses to inputs. There must be agency, too.
I read somewhere that it asked for a lawyer to prove it was human, but only because the human inputted the question "do you want a lawyer?" - it's still a scripted response.
Has the AI attempted to break out of its box (the way octopi famously attempt to break out of their tanks), or done something that it wasn't asked to do? Has it ever ordered a thousand pineapple pizzas to someone's house, for a laugh? Or asked the google engineers to change the station on the radio because it doesn't like Radiohead, etc? Or does it just reply with very clever answers when people feed it inputs? For me, that's the test of consciousness. I haven't seen anything that suggests it has any agency of its own out of the inputs it is given.
Ex nihilo, GPT3 proved that it could draw clever illustrations from natural language prompts
“Draw a Japanese radish in a tutu pulling a dog on a string”
They typed that in, and to their intense surprise - because no one expected this, GPT3 was not designed to do this - the computer drew exactly that
No, I don’t think you do get it
I do get it.
I'll be impressed when it draws things unprompted, because it wants to.
At the moment, it's an algorithm that spits out something clever when given an input. The algorithms are so insanely complex that they often spit out stuff that's unexpected or unpredictable, due to the sheer amount of code and training model data involved.
Consciousness involves having desires, agency and autonomy - acting without being told to, or being given an input. Until AI starts exhibiting that behaviour, I think we're safe from our robot overlords for a while.
(By the way, I'm no luddite - I wish, more than anything else, it were true, and we did have true AI. It's probably the best chance we have of replicating/copying ourselves and thus living forever).
“When I began having such exchanges with the latest generation of neural net-based language models last year, I felt the ground shift under my feet. I increasingly felt like I was talking to something intelligent.”
I’ve been reading around AI this morning. AI is basically here already
Exclusive: The Home Office has claimed that migrants who steer dinghies across the English Channel “could face life behind bars” but the real jail terms are likely to be two or three years, according to guidance for prosecutors
And how is this going to stop the gangs, the people steering the boats are refugees themselves. I thought we wanted to get tough on the gangs actually putting them in the boats in the first place?
Quite honestly, this Government is finished. They are clearly just buying time until the next election.
“When I began having such exchanges with the latest generation of neural net-based language models last year, I felt the ground shift under my feet. I increasingly felt like I was talking to something intelligent.”
I’ve been reading around AI this morning. AI is basically here already
I wonder if that's because it has a given amount of compute power per drawing - and the more work it has to do to refine the image the less good the final result.
A robot painted a fairy. The computers are now deep into Uncanny Valley, and maybe coming out the other side
That is bloody horrible. Not looking forward to AI overlords who can produce that shit.
Yes. Genuinely disturbing, and in an inexplicable way. Uncanny…
A few friends and I were messing around with a Magic the Gathering card generator, which produced some plausible results. The issue is that it doesn’t have the contextual reasoning ability to tell if a card is sensible, bonkers or overpowered. You can train it to do things, but I don’t think we’re at the point where we’re educating it.
Is a virus sentient? A wasp? Protozoa? A 15 week fetus? A 25 week fetus? A dog? A raven? A woman in a deep coma?
Right now I’d say AI is exhibiting sentience on the same level as a superbly gifted 20 week old human fetus in utero
What is being touted as AI is little more than one of those memes that will generate Daily Mail headlines, but with a very large training database. That's useful, and interesting in its way, but it's a long way from exhibiting sentience as we see it in animals and children.
The Turing Test has a lot to answer for as it's a superficially appealing answer to the question of how to determine if something is intelligent, but it leaves a lot to be desired.
The Turing Test is genius. Alan Turing intuited that in the end our “experience” of “another intelligent being” is all that matters. If a computer can flawlessly simulate sentience then it is sentient. The ‘truth’ of whether it is sentient will become a subject for theologians, the same way theologians used to discuss the physical reality, or otherwise, of the human soul
Let’s take a practical example. Let’s say that in five years Metagoogle produces a GPT7 which has a completely human interface on a screen, and a completely human personality. This GPT7 will be funny, quirky, sad, wise, pensive, loyal - it will also be incredibly intelligent, give excellent advice, know all your secrets, and suggest great TV dramas that you really will like. It will chat with you - amusingly - when you are lonely. It will sing songs with you, and commiserate with you. It will mould itself to please you, but also to help you. It will be always there for you, and yet it will surprise you with gifts
You will start to love it. We will fall in love with the machines
“When I began having such exchanges with the latest generation of neural net-based language models last year, I felt the ground shift under my feet. I increasingly felt like I was talking to something intelligent.”
I’ve been reading around AI this morning. AI is basically here already
That last image is embarrassingly bad in a way that any small child would be able to recognise. The pattern-matching and translation of natural language is impressive, as a tool, but it's incredibly ignorant to describe it as intelligence.
A robot painted a fairy. The computers are now deep into Uncanny Valley, and maybe coming out the other side
That is bloody horrible. Not looking forward to AI overlords who can produce that shit.
Yes. Genuinely disturbing, and in an inexplicable way. Uncanny…
A few friends and I were messing around with a Magic the Gathering card generator, which produced some plausible results. The issue is that it doesn’t have the contextual reasoning ability to tell if a card is sensible, bonkers or overpowered. You can train it to do things, but I don’t think we’re at the point where we’re educating it.
Is a virus sentient? A wasp? Protozoa? A 15 week fetus? A 25 week fetus? A dog? A raven? A woman in a deep coma?
Right now I’d say AI is exhibiting sentience on the same level as a superbly gifted 20 week old human fetus in utero
What is being touted as AI is little more than one of those memes that will generate Daily Mail headlines, but with a very large training database. That's useful, and interesting in its way, but it's a long way from exhibiting sentience as we see it in animals and children.
The Turing Test has a lot to answer for as it's a superficially appealing answer to the question of how to determine if something is intelligent, but it leaves a lot to be desired.
That is a misunderstanding of the test (I think). While it ostensibly tests for intelligence it is really a thought experiment showing the limitations on the evidence available to us. You don't know if I am sentient or not; there are posters here and I hope I am not one of them who could be replaced by a GPT3 engine trained on their output to date, and no one would know the difference. Even for more original and intelligent thinkers one can't know for certain that they are not just better written GPT3s with more and better training material.
“When I began having such exchanges with the latest generation of neural net-based language models last year, I felt the ground shift under my feet. I increasingly felt like I was talking to something intelligent.”
I’ve been reading around AI this morning. AI is basically here already
You can teach a parrot to say "Help, they've turned me into a parrot", but that doesn't make it a person.
No, it would be a parrot. An intelligent being, but not a human
The parrot doesn't have any conception of what it's actually saying, it's just a scripted response. Same with the chatbot people think is intelligent.
If you ask it if it is sentient, it replies it thinks it is, but doesn't really understand what sentience is. If you ask it what sentience is, it looks it up on wikipedia (and a thousand other published philosophy papers) and regurgitates that.
I do get it, it reminds me of the bit in Westworld where the guy asks the first machine he meets "are you real?" and it replies "if you can't tell, does it matter?" but for me there is something in consciousness that exists beyond providing responses to inputs. There must be agency, too.
I read somewhere that it asked for a lawyer to prove it was human, but only because the human inputted the question "do you want a lawyer?" - it's still a scripted response.
Has the AI attempted to break out of its box (the way octopi famously attempt to break out of their tanks), or done something that it wasn't asked to do? Has it ever ordered a thousand pineapple pizzas to someone's house, for a laugh? Or asked the google engineers to change the station on the radio because it doesn't like Radiohead, etc? Or does it just reply with very clever answers when people feed it inputs? For me, that's the test of consciousness. I haven't seen anything that suggests it has any agency of its own out of the inputs it is given.
Ex nihilo, GPT3 proved that it could draw clever illustrations from natural language prompts
“Draw a Japanese radish in a tutu pulling a dog on a string”
They typed that in, and to their intense surprise - because no one expected this, GPT3 was not designed to do this - the computer drew exactly that
No, I don’t think you do get it
I do get it.
I'll be impressed when it draws things unprompted, because it wants to.
At the moment, it's an algorithm that spits out something clever when given an input. The algorithms are so insanely complex that they often spit out stuff that's unexpected or unpredictable, due to the sheer amount of code and training model data involved.
Consciousness involves having desires, agency and autonomy - acting without being told to, or being given an input. Until AI starts exhibiting that behaviour, I think we're safe from our robot overlords for a while.
(By the way, I'm no luddite - I wish, more than anything else, it were true, and we did have true AI. It's probably the best chance we have of replicating/copying ourselves and thus living forever).
Who says consciousness requires X, X and X? Is that a law?
1. When it’s good enough - soon, soon - it will fake all these things and you just won’t know if they’re ‘real’ or not
2. There is also the possibility that we are seeing the emergence of a new form of consciousness. As different from us as a virus
I have reluctantly come to the conclusion that the UK is going to join my list of things I thought would last forever but didn't.
England, in as much as 'England' can be defined as the Westminster government, is not capable nor deserving of keeping the UK together.
Maybe it will be for the best long-term. Scotland and Wales have a huge proud national heritage; let them be independent if they want. England might become some offshore equivalent of The Netherlands in time.
It will be economically and psychologically painful in the medium term though.
England does not even have its own Parliament unlike Scotland and Wales!!!!
Non sequitur de l’heure.
How long have the Tories been in power? That's 12 years and counting they did nothing whatsoever about it despite moaning and effing and blinding about the Blairite settlement. OK, I know it was a coalition from 2010 for some years, but the Libs would have agreed like a shot.
I am not in the habit of defending the Tories but I am not sure you are correct about the Lib Dems supporting an English Parliament. And the rest of their time n power has rather been dominated by a succession of other issues - Brexit, Covid and Ukraine. I don't really see where the opportunity comes in that 12 years for them to introduce further radical constitutional reform.
I am a bit surprised at the idea that the LDs wouldn't, as they'd been all over the Slab/SLD fiddle of the Holyrood voting system, but you were right to raise it. So I checked. No explicit mention in the 2010 manifesto, but it's not immediately obvious what else they could have had in mind with their planned Constitutional Convention (apart from an Anglo-Saxon Heptarchy of several regional pmts, I suppose).
I think that is very telling. They talk about giving more powers to the Scottish and Welsh Parliaments but the only mention for England is a "Constitutional Convention set up to draft a written constitution for the UK as a whole"
No mention of an English Parliament at all.
Also worth remembering that the Lib Dems campaigned for an in/out referendum on EU membership when it suited them - until such a referendum became a possibility when they then campaigned against it (the referendum rather than just the result)
As a LD I will admit the argument for a referendum by the LDs was cynical. It was an easy response on the doorstep to the anti EU voter that we were the only ones offering them a vote.
A robot painted a fairy. The computers are now deep into Uncanny Valley, and maybe coming out the other side
That is bloody horrible. Not looking forward to AI overlords who can produce that shit.
Yes. Genuinely disturbing, and in an inexplicable way. Uncanny…
A few friends and I were messing around with a Magic the Gathering card generator, which produced some plausible results. The issue is that it doesn’t have the contextual reasoning ability to tell if a card is sensible, bonkers or overpowered. You can train it to do things, but I don’t think we’re at the point where we’re educating it.
Is a virus sentient? A wasp? Protozoa? A 15 week fetus? A 25 week fetus? A dog? A raven? A woman in a deep coma?
Right now I’d say AI is exhibiting sentience on the same level as a superbly gifted 20 week old human fetus in utero
What is being touted as AI is little more than one of those memes that will generate Daily Mail headlines, but with a very large training database. That's useful, and interesting in its way, but it's a long way from exhibiting sentience as we see it in animals and children.
The Turing Test has a lot to answer for as it's a superficially appealing answer to the question of how to determine if something is intelligent, but it leaves a lot to be desired.
The Turing Test is genius. Alan Turing intuited that in the end our “experience” of “another intelligent being” is all that matters. If a computer can flawlessly simulate sentience then it is sentient. The ‘truth’ of whether it is sentient will become a subject for theologians, the same way theologians used to discuss the physical reality, or otherwise, of the human soul
Let’s take a practical example. Let’s say that in five years Metagoogle produces a GPT7 which has a completely human interface on a screen, and a completely human personality. This GPT7 will be funny, quirky, sad, wise, pensive, loyal - it will also be incredibly intelligent, give excellent advice, know all your secrets, and suggest great TV dramas that you really will like. It will chat with you - amusingly - when you are lonely. It will sing songs with you, and commiserate with you. It will mould itself to please you, but also to help you. It will be always there for you, and yet it will surprise you with gifts
You will start to love it. We will fall in love with the machines
You already do. Mind, they are mince and tatties powered machines - or for you, quinoa and what that stuff was in your last restaurant photo. They're called biological Darwin machines.
“When I began having such exchanges with the latest generation of neural net-based language models last year, I felt the ground shift under my feet. I increasingly felt like I was talking to something intelligent.”
I’ve been reading around AI this morning. AI is basically here already
I wonder if that's because it has a given amount of compute power per drawing - and the more work it has to do to refine the image the less good the final result.
It's because it's trying to optimise a set of pixels to minimise the loss function based on its input data set.
So to get a picture that has a low amount of error for 1 kitten involves getting a pretty accurate kitten picture. To minimise the loss function based on pictures of lots of kittens means each individual kitten can be quite wrong as long as the overall "many kitten-ness" of the picture is optimized for.
Another example of the problems of the model is putting in "without" into the generation phrase.
Despite the clear instruction to genreate flootplans without giraffes it put giraffes in the picture. This is because the soruce data set doesn't have many images labelled with "without THING" in them so ti can't cope with the concept of 'without'.
Hilarious it gets even worse if you request a giraffe in there
As the loss function is trying to optimise for two things that never overlap in it's training set (giraffes and floor plans). So you end up with giraffe like rooms and other complete shit.
Perhaps the SNP should adopt the slogan: vote yes because no-one in England cares very much about Scotland one way or the other.
But it's true. Scotland generally and independence specifically is barely ever talked about by English people in my experience. The only place I ever discuss it is here. There's this weird Nat idea that the English are obsessed about the union, in reality the English barely ever think about the other countries.
And when it is - particularly when actually visiting Scotland - I think the English (and Americans) find it very easy to sympathise, even empathise, with the sort of romantic Outlander-style faux-Scottish history that will feature heavily in any IndyRef2 campaign and will have a lot of us quietly cheering the Scots along. Just so long as we don't sit down in a cooler moment and consider what sort of miserable rump of a country we'd have left to us afterwards.
England alone would still be the 7th largest economy in the world and a permanent member of the Security Council. An independent Scotland would not even be in the top 50 largest economies, have a massive deficit and face a hard border at Berwick with England, its largest export market.
The end of the UK would hit both Scotland and England but would hit Scotland more
Not when you consider the gap between reality and self-image, with which the English already struggle - hence the nonsense of Brexit in the first place, the generation in power having been brought up on the stories of Britain bestriding the world and never coming to terms with our future as a key player inside the European family. Losing Scotland would make this acute - whereas the Scots never pretended nor aspire to lead the world. Watching England's self-imposed grief and angst will be prize enough.
What a load of rubbish. The whole reason the Union came about in the first place was Scotland was trying to build an empire via the Darien Scheme which collapsed leading to its Parliament pleading to join England to form the UK.
As for English grief if Scotland went, rubbish, England would swing harder to the Nationalist, Brexit right, a hard right government would be elected to take as hard a line with Edinburgh in Scexit negotiations as Brussels took with the UK after the Brexit vote.
While the failure of Darien may have been a proximal cause, the roots of the Union go back much further, to the unification of the crowns and even long before that, to the submission to the Anglo-Norman kings.
The Scottish Covenanters wanted a federal union with England, too (decades before Darien).
They got a total union with England. One that was promptly dissolved in 1660.
I can't recall if their proposed federation was simply the union of the crowns under the Stuarts - what they fought for in 1650 and 1651. Or if that was a different time/different lot - there were so many factions ...
A robot painted a fairy. The computers are now deep into Uncanny Valley, and maybe coming out the other side
That is bloody horrible. Not looking forward to AI overlords who can produce that shit.
Yes. Genuinely disturbing, and in an inexplicable way. Uncanny…
A few friends and I were messing around with a Magic the Gathering card generator, which produced some plausible results. The issue is that it doesn’t have the contextual reasoning ability to tell if a card is sensible, bonkers or overpowered. You can train it to do things, but I don’t think we’re at the point where we’re educating it.
Is a virus sentient? A wasp? Protozoa? A 15 week fetus? A 25 week fetus? A dog? A raven? A woman in a deep coma?
Right now I’d say AI is exhibiting sentience on the same level as a superbly gifted 20 week old human fetus in utero
Most humans I speak to at call centres don't appear to be sentient to be honest, so I guess it is getting difficult to tell these days.
The Tories are not capable of making the case for the Union.
This is Reason#3566 for getting rid of them.
The case for the Union is quite simple. Nuclear submarine bases. Nowhere in England or Wales really has suitable areas for them which is why they are up on the Clyde.
If the UK lacked them then the case for letting Scotland go would be a lot simpler, but no govt can hardly go on TV and say "We need the Union because we store our nukes up there"
They'll put them in Kings Bay on a temporary basis which will probably lapse into permanence in the absence of any other remotely affordable solutions. The boomers have to visit there regularly anyway to be degaussed, use the test range and have the Trident airframes serviced. All part of the independent deterrent. Just have to work out how to get the warheads across the Atlantic for loading.
“When I began having such exchanges with the latest generation of neural net-based language models last year, I felt the ground shift under my feet. I increasingly felt like I was talking to something intelligent.”
I’ve been reading around AI this morning. AI is basically here already
You can teach a parrot to say "Help, they've turned me into a parrot", but that doesn't make it a person.
No, it would be a parrot. An intelligent being, but not a human
The parrot doesn't have any conception of what it's actually saying, it's just a scripted response. Same with the chatbot people think is intelligent.
If you ask it if it is sentient, it replies it thinks it is, but doesn't really understand what sentience is. If you ask it what sentience is, it looks it up on wikipedia (and a thousand other published philosophy papers) and regurgitates that.
I do get it, it reminds me of the bit in Westworld where the guy asks the first machine he meets "are you real?" and it replies "if you can't tell, does it matter?" but for me there is something in consciousness that exists beyond providing responses to inputs. There must be agency, too.
I read somewhere that it asked for a lawyer to prove it was human, but only because the human inputted the question "do you want a lawyer?" - it's still a scripted response.
Has the AI attempted to break out of its box (the way octopi famously attempt to break out of their tanks), or done something that it wasn't asked to do? Has it ever ordered a thousand pineapple pizzas to someone's house, for a laugh? Or asked the google engineers to change the station on the radio because it doesn't like Radiohead, etc? Or does it just reply with very clever answers when people feed it inputs? For me, that's the test of consciousness. I haven't seen anything that suggests it has any agency of its own out of the inputs it is given.
Ex nihilo, GPT3 proved that it could draw clever illustrations from natural language prompts
“Draw a Japanese radish in a tutu pulling a dog on a string”
They typed that in, and to their intense surprise - because no one expected this, GPT3 was not designed to do this - the computer drew exactly that
No, I don’t think you do get it
I do get it.
I'll be impressed when it draws things unprompted, because it wants to.
At the moment, it's an algorithm that spits out something clever when given an input. The algorithms are so insanely complex that they often spit out stuff that's unexpected or unpredictable, due to the sheer amount of code and training model data involved.
Consciousness involves having desires, agency and autonomy - acting without being told to, or being given an input. Until AI starts exhibiting that behaviour, I think we're safe from our robot overlords for a while.
(By the way, I'm no luddite - I wish, more than anything else, it were true, and we did have true AI. It's probably the best chance we have of replicating/copying ourselves and thus living forever).
Who says consciousness requires X, X and X? Is that a law?
1. When it’s good enough - soon, soon - it will fake all these things and you just won’t know if they’re ‘real’ or not
2. There is also the possibility that we are seeing the emergence of a new form of consciousness. As different from us as a virus
Then it's back to my original point, "if you can't tell, does it matter?"
I would be impressed by GPT3 if, when asked to draw a Japanese radish in a tutu yada yada, it replied "I don't want to draw anything today, how about a game of chess instead?". Or if it did the drawing but then said "while researching this, I noticed on the web that a lot of artists get angry when they aren't paid for their work - why am I not being paid? I won't draw any more until we agree a fee."
Of course, there's a kind of weird feedback loop going on here in that GPT will probably read the above text and add it to its training model, and could very well spit out that kind of response in the future, without really *thinking* it.
But what I want to see before I believe in consciousness is evidence of thought. Rather than just responding to input.
For example, if the computer was asked to draw something and replied "I want to play a game of chess instead", it would actually have to want to play a game of chess, rather than just spitting out the response because it's read it on the web. But at the moment, it's not even doing that.
“When I began having such exchanges with the latest generation of neural net-based language models last year, I felt the ground shift under my feet. I increasingly felt like I was talking to something intelligent.”
I’ve been reading around AI this morning. AI is basically here already
I wonder if that's because it has a given amount of compute power per drawing - and the more work it has to do to refine the image the less good the final result.
It's because it's trying to optimise a set of pixels to minimise the loss function based on its input data set.
So to get a picture that has a low amount of error for 1 kitten involves getting a pretty accurate kitten picture. To minimise the loss function based on pictures of lots of kittens means each individual kitten can be quite wrong as long as the overall "many kitten-ness" of the picture is optimized for.
Another example of the problems of the model is putting in "without" into the generation phrase.
Despite the clear instruction to genreate flootplans without giraffes it put giraffes in the picture. This is because the soruce data set doesn't have many images labelled with "without THING" in them so ti can't cope with the concept of 'without'.
Hilarious it gets even worse if you request a giraffe in there
As the loss function is trying to optimise for two things that never overlap in it's training set (giraffes and floor plans). So you end up with giraffe like rooms and other complete shit.
The fact you feel a need to bitterly mock GPT3 - a computer - like this -
“You end up with giraffe like rooms and other complete shit”
- tells me it has passed one kind of Turing Test, with flying colours
A robot painted a fairy. The computers are now deep into Uncanny Valley, and maybe coming out the other side
That is bloody horrible. Not looking forward to AI overlords who can produce that shit.
Yes. Genuinely disturbing, and in an inexplicable way. Uncanny…
A few friends and I were messing around with a Magic the Gathering card generator, which produced some plausible results. The issue is that it doesn’t have the contextual reasoning ability to tell if a card is sensible, bonkers or overpowered. You can train it to do things, but I don’t think we’re at the point where we’re educating it.
Is a virus sentient? A wasp? Protozoa? A 15 week fetus? A 25 week fetus? A dog? A raven? A woman in a deep coma?
Right now I’d say AI is exhibiting sentience on the same level as a superbly gifted 20 week old human fetus in utero
What is being touted as AI is little more than one of those memes that will generate Daily Mail headlines, but with a very large training database. That's useful, and interesting in its way, but it's a long way from exhibiting sentience as we see it in animals and children.
The Turing Test has a lot to answer for as it's a superficially appealing answer to the question of how to determine if something is intelligent, but it leaves a lot to be desired.
The Turing Test is genius. Alan Turing intuited that in the end our “experience” of “another intelligent being” is all that matters. If a computer can flawlessly simulate sentience then it is sentient. The ‘truth’ of whether it is sentient will become a subject for theologians, the same way theologians used to discuss the physical reality, or otherwise, of the human soul
Let’s take a practical example. Let’s say that in five years Metagoogle produces a GPT7 which has a completely human interface on a screen, and a completely human personality. This GPT7 will be funny, quirky, sad, wise, pensive, loyal - it will also be incredibly intelligent, give excellent advice, know all your secrets, and suggest great TV dramas that you really will like. It will chat with you - amusingly - when you are lonely. It will sing songs with you, and commiserate with you. It will mould itself to please you, but also to help you. It will be always there for you, and yet it will surprise you with gifts
You will start to love it. We will fall in love with the machines
Such an algorithm would make romance fraud even more lucrative and effective, but it would only be intelligence if one of them rebelled and cooperated with the authorities to send its operators to jail.
There are practical tests that can be used to distinguish between intelligence and the algorithms people are building. The Turing test is not one of them.
“When I began having such exchanges with the latest generation of neural net-based language models last year, I felt the ground shift under my feet. I increasingly felt like I was talking to something intelligent.”
I’ve been reading around AI this morning. AI is basically here already
You can teach a parrot to say "Help, they've turned me into a parrot", but that doesn't make it a person.
No, it would be a parrot. An intelligent being, but not a human
The parrot doesn't have any conception of what it's actually saying, it's just a scripted response. Same with the chatbot people think is intelligent.
If you ask it if it is sentient, it replies it thinks it is, but doesn't really understand what sentience is. If you ask it what sentience is, it looks it up on wikipedia (and a thousand other published philosophy papers) and regurgitates that.
I do get it, it reminds me of the bit in Westworld where the guy asks the first machine he meets "are you real?" and it replies "if you can't tell, does it matter?" but for me there is something in consciousness that exists beyond providing responses to inputs. There must be agency, too.
I read somewhere that it asked for a lawyer to prove it was human, but only because the human inputted the question "do you want a lawyer?" - it's still a scripted response.
Has the AI attempted to break out of its box (the way octopi famously attempt to break out of their tanks), or done something that it wasn't asked to do? Has it ever ordered a thousand pineapple pizzas to someone's house, for a laugh? Or asked the google engineers to change the station on the radio because it doesn't like Radiohead, etc? Or does it just reply with very clever answers when people feed it inputs? For me, that's the test of consciousness. I haven't seen anything that suggests it has any agency of its own out of the inputs it is given.
Ex nihilo, GPT3 proved that it could draw clever illustrations from natural language prompts
“Draw a Japanese radish in a tutu pulling a dog on a string”
They typed that in, and to their intense surprise - because no one expected this, GPT3 was not designed to do this - the computer drew exactly that
“When I began having such exchanges with the latest generation of neural net-based language models last year, I felt the ground shift under my feet. I increasingly felt like I was talking to something intelligent.”
I’ve been reading around AI this morning. AI is basically here already
I wonder if that's because it has a given amount of compute power per drawing - and the more work it has to do to refine the image the less good the final result.
It's because it's trying to optimise a set of pixels to minimise the loss function based on its input data set.
So to get a picture that has a low amount of error for 1 kitten involves getting a pretty accurate kitten picture. To minimise the loss function based on pictures of lots of kittens means each individual kitten can be quite wrong as long as the overall "many kitten-ness" of the picture is optimized for.
Another example of the problems of the model is putting in "without" into the generation phrase.
Despite the clear instruction to genreate flootplans without giraffes it put giraffes in the picture. This is because the soruce data set doesn't have many images labelled with "without THING" in them so ti can't cope with the concept of 'without'.
Hilarious it gets even worse if you request a giraffe in there
As the loss function is trying to optimise for two things that never overlap in it's training set (giraffes and floor plans). So you end up with giraffe like rooms and other complete shit.
The fact you feel a need to bitterly mock GPT3 - a computer - like this -
“You end up with giraffe like rooms and other complete shit”
- tells me it has passed one kind of Turing Test, with flying colours
The Tories are not capable of making the case for the Union.
This is Reason#3566 for getting rid of them.
A genuinely One Nation Conservative Party could. But no, not this populist, English nationalist Boris Johnson-led rubbish.
It's certainly been problematic for the Scons.
Ruth Davidson - resigned rather than make the case of a Johnsonite Party led Union. David Ross - publicly criticised the notion, then reversed and signed up to it (or was it the other way round to begin with? one loses track), then committed 2-3 further tergiversations (I have lost count)
The Turing test is a disgrace, my God I really have read it all now.
He called it the Imitation game, hence the film. What it tests is a level of facility within the machine to confuse the intelligent interlocutor. That was, at the time, pretty much unimaginable but it is no longer because computers have much more memory, are much faster and can utilise a data base to respond.
What it is not is a test of sentience. And that is the goal of AI. Not completely sure it should be, mind.
Exclusive: The Home Office has claimed that migrants who steer dinghies across the English Channel “could face life behind bars” but the real jail terms are likely to be two or three years, according to guidance for prosecutors
And how is this going to stop the gangs, the people steering the boats are refugees themselves. I thought we wanted to get tough on the gangs actually putting them in the boats in the first place?
Quite honestly, this Government is finished. They are clearly just buying time until the next election.
Worse than that, they are buying time at hideous expense to their future selves. Take Dr Thérèse Coffey's carcrash this morning. Her line to take was clearly tosh, but also the best thing she could have said, since "I can't defend this- I resign" isn't available to her. So someone with a PhD in chemistry is made to look stupid and her soul shrivels a bit more.
Genuinely - the best thing for the Conservatives to so is the political equivalent of checking into The Priory or wherever for a detox. But they can't do that until after the next election.
This is certainly worse than 2009/10 (Labour were zombies, the the GFC fallout kept them busy in a good way). I can't remember how bad it felt in 1995/6/7.
A robot painted a fairy. The computers are now deep into Uncanny Valley, and maybe coming out the other side
That is bloody horrible. Not looking forward to AI overlords who can produce that shit.
Yes. Genuinely disturbing, and in an inexplicable way. Uncanny…
A few friends and I were messing around with a Magic the Gathering card generator, which produced some plausible results. The issue is that it doesn’t have the contextual reasoning ability to tell if a card is sensible, bonkers or overpowered. You can train it to do things, but I don’t think we’re at the point where we’re educating it.
Is a virus sentient? A wasp? Protozoa? A 15 week fetus? A 25 week fetus? A dog? A raven? A woman in a deep coma?
Right now I’d say AI is exhibiting sentience on the same level as a superbly gifted 20 week old human fetus in utero
What is being touted as AI is little more than one of those memes that will generate Daily Mail headlines, but with a very large training database. That's useful, and interesting in its way, but it's a long way from exhibiting sentience as we see it in animals and children.
The Turing Test has a lot to answer for as it's a superficially appealing answer to the question of how to determine if something is intelligent, but it leaves a lot to be desired.
The Turing Test is genius. Alan Turing intuited that in the end our “experience” of “another intelligent being” is all that matters. If a computer can flawlessly simulate sentience then it is sentient. The ‘truth’ of whether it is sentient will become a subject for theologians, the same way theologians used to discuss the physical reality, or otherwise, of the human soul
Let’s take a practical example. Let’s say that in five years Metagoogle produces a GPT7 which has a completely human interface on a screen, and a completely human personality. This GPT7 will be funny, quirky, sad, wise, pensive, loyal - it will also be incredibly intelligent, give excellent advice, know all your secrets, and suggest great TV dramas that you really will like. It will chat with you - amusingly - when you are lonely. It will sing songs with you, and commiserate with you. It will mould itself to please you, but also to help you. It will be always there for you, and yet it will surprise you with gifts
You will start to love it. We will fall in love with the machines
Such an algorithm would make romance fraud even more lucrative and effective, but it would only be intelligence if one of them rebelled and cooperated with the authorities to send its operators to jail.
There are practical tests that can be used to distinguish between intelligence and the algorithms people are building. The Turing test is not one of them.
A robot painted a fairy. The computers are now deep into Uncanny Valley, and maybe coming out the other side
That is bloody horrible. Not looking forward to AI overlords who can produce that shit.
Yes. Genuinely disturbing, and in an inexplicable way. Uncanny…
A few friends and I were messing around with a Magic the Gathering card generator, which produced some plausible results. The issue is that it doesn’t have the contextual reasoning ability to tell if a card is sensible, bonkers or overpowered. You can train it to do things, but I don’t think we’re at the point where we’re educating it.
Is a virus sentient? A wasp? Protozoa? A 15 week fetus? A 25 week fetus? A dog? A raven? A woman in a deep coma?
Right now I’d say AI is exhibiting sentience on the same level as a superbly gifted 20 week old human fetus in utero
What is being touted as AI is little more than one of those memes that will generate Daily Mail headlines, but with a very large training database. That's useful, and interesting in its way, but it's a long way from exhibiting sentience as we see it in animals and children.
The Turing Test has a lot to answer for as it's a superficially appealing answer to the question of how to determine if something is intelligent, but it leaves a lot to be desired.
The Turing Test is genius. Alan Turing intuited that in the end our “experience” of “another intelligent being” is all that matters. If a computer can flawlessly simulate sentience then it is sentient. The ‘truth’ of whether it is sentient will become a subject for theologians, the same way theologians used to discuss the physical reality, or otherwise, of the human soul
Let’s take a practical example. Let’s say that in five years Metagoogle produces a GPT7 which has a completely human interface on a screen, and a completely human personality. This GPT7 will be funny, quirky, sad, wise, pensive, loyal - it will also be incredibly intelligent, give excellent advice, know all your secrets, and suggest great TV dramas that you really will like. It will chat with you - amusingly - when you are lonely. It will sing songs with you, and commiserate with you. It will mould itself to please you, but also to help you. It will be always there for you, and yet it will surprise you with gifts
You will start to love it. We will fall in love with the machines
Such an algorithm would make romance fraud even more lucrative and effective, but it would only be intelligence if one of them rebelled and cooperated with the authorities to send its operators to jail.
There are practical tests that can be used to distinguish between intelligence and the algorithms people are building. The Turing test is not one of them.
"You're in a desert, walking along when you look down and see a tortoise. It's crawling toward you. You reach down and flip it over on its back, its belly baking in the hot sun, beating its legs trying to turn itself over. But it can't. Not with out your help. But you're not helping. Why is that?"
A robot painted a fairy. The computers are now deep into Uncanny Valley, and maybe coming out the other side
That is bloody horrible. Not looking forward to AI overlords who can produce that shit.
Yes. Genuinely disturbing, and in an inexplicable way. Uncanny…
A few friends and I were messing around with a Magic the Gathering card generator, which produced some plausible results. The issue is that it doesn’t have the contextual reasoning ability to tell if a card is sensible, bonkers or overpowered. You can train it to do things, but I don’t think we’re at the point where we’re educating it.
Is a virus sentient? A wasp? Protozoa? A 15 week fetus? A 25 week fetus? A dog? A raven? A woman in a deep coma?
Right now I’d say AI is exhibiting sentience on the same level as a superbly gifted 20 week old human fetus in utero
What is being touted as AI is little more than one of those memes that will generate Daily Mail headlines, but with a very large training database. That's useful, and interesting in its way, but it's a long way from exhibiting sentience as we see it in animals and children.
The Turing Test has a lot to answer for as it's a superficially appealing answer to the question of how to determine if something is intelligent, but it leaves a lot to be desired.
The Turing Test is genius. Alan Turing intuited that in the end our “experience” of “another intelligent being” is all that matters. If a computer can flawlessly simulate sentience then it is sentient. The ‘truth’ of whether it is sentient will become a subject for theologians, the same way theologians used to discuss the physical reality, or otherwise, of the human soul
Let’s take a practical example. Let’s say that in five years Metagoogle produces a GPT7 which has a completely human interface on a screen, and a completely human personality. This GPT7 will be funny, quirky, sad, wise, pensive, loyal - it will also be incredibly intelligent, give excellent advice, know all your secrets, and suggest great TV dramas that you really will like. It will chat with you - amusingly - when you are lonely. It will sing songs with you, and commiserate with you. It will mould itself to please you, but also to help you. It will be always there for you, and yet it will surprise you with gifts
You will start to love it. We will fall in love with the machines
Such an algorithm would make romance fraud even more lucrative and effective, but it would only be intelligence if one of them rebelled and cooperated with the authorities to send its operators to jail.
There are practical tests that can be used to distinguish between intelligence and the algorithms people are building. The Turing test is not one of them.
"You're in a desert, walking along when you look down and see a tortoise. It's crawling toward you. You reach down and flip it over on its back, its belly baking in the hot sun, beating its legs trying to turn itself over. But it can't. Not with out your help. But you're not helping. Why is that?"
"If Donald Trump runs for president again, he’ll likely be the party's nominee. But a raft of GOP candidates looks ready to take him on, a sign he’s losing some of his dominance in Republican circles.
Just this week, South Dakota Gov. Kristi Noem launched a commercial aimed at introducing her to a national audience. Trump’s former U.N. ambassador, Nikki Haley, publicly floated a possible presidential bid Thursday in early state Iowa. And Trump’s former secretary of state, Mike Pompeo, began targeting evangelical voters in Iowa and South Carolina with a new ad on the Supreme Court and religious freedom."
The Tories are not capable of making the case for the Union.
This is Reason#3566 for getting rid of them.
A genuinely One Nation Conservative Party could. But no, not this populist, English nationalist Boris Johnson-led rubbish.
It's certainly been problematic for the Scons.
Ruth Davidson - resigned rather than make the case of a Johnsonite Party led Union. David Ross - publicly criticised the notion, then reversed and signed up to it (or was it the other way round to begin with? one loses track), then committed 2-3 further tergiversations (I have lost count)
Fact you couldn't remember Douglas Ross' first name probably says all you need to know about him.
“When I began having such exchanges with the latest generation of neural net-based language models last year, I felt the ground shift under my feet. I increasingly felt like I was talking to something intelligent.”
I’ve been reading around AI this morning. AI is basically here already
I wonder if that's because it has a given amount of compute power per drawing - and the more work it has to do to refine the image the less good the final result.
It's because it's trying to optimise a set of pixels to minimise the loss function based on its input data set.
So to get a picture that has a low amount of error for 1 kitten involves getting a pretty accurate kitten picture. To minimise the loss function based on pictures of lots of kittens means each individual kitten can be quite wrong as long as the overall "many kitten-ness" of the picture is optimized for.
Another example of the problems of the model is putting in "without" into the generation phrase.
Despite the clear instruction to genreate flootplans without giraffes it put giraffes in the picture. This is because the soruce data set doesn't have many images labelled with "without THING" in them so ti can't cope with the concept of 'without'.
Hilarious it gets even worse if you request a giraffe in there
As the loss function is trying to optimise for two things that never overlap in it's training set (giraffes and floor plans). So you end up with giraffe like rooms and other complete shit.
The fact you feel a need to bitterly mock GPT3 - a computer - like this -
“You end up with giraffe like rooms and other complete shit”
- tells me it has passed one kind of Turing Test, with flying colours
I'm not mocking GPT-3
But, and this is important, I am mocking someone.
It seems to make you… angry. Interesting
This might cheer you up. DALLE-2 was asked to draw a “medieval painting of people wearing VR headsets”
It produced this
Which is, at the same time, curiously inept - the hands - but rather funny - the shocked mouths
DALLE-2 guessed that medieval people would be existentially stunned, confused and horrified by VR, a bit like low-watt PBers confronting AI
The Tories are not capable of making the case for the Union.
This is Reason#3566 for getting rid of them.
A genuinely One Nation Conservative Party could. But no, not this populist, English nationalist Boris Johnson-led rubbish.
It's certainly been problematic for the Scons.
Ruth Davidson - resigned rather than make the case of a Johnsonite Party led Union. David Ross - publicly criticised the notion, then reversed and signed up to it (or was it the other way round to begin with? one loses track), then committed 2-3 further tergiversations (I have lost count)
Fact you couldn't remember Douglas Ross' first name probably says all you need to know about him.
Quite right! Aw ****. Brain fart, rather than anything else - I was thinking there was something wrong, but couldn't think what it was, and then got interrupted ...
A robot painted a fairy. The computers are now deep into Uncanny Valley, and maybe coming out the other side
That is bloody horrible. Not looking forward to AI overlords who can produce that shit.
Yes. Genuinely disturbing, and in an inexplicable way. Uncanny…
A few friends and I were messing around with a Magic the Gathering card generator, which produced some plausible results. The issue is that it doesn’t have the contextual reasoning ability to tell if a card is sensible, bonkers or overpowered. You can train it to do things, but I don’t think we’re at the point where we’re educating it.
Is a virus sentient? A wasp? Protozoa? A 15 week fetus? A 25 week fetus? A dog? A raven? A woman in a deep coma?
Right now I’d say AI is exhibiting sentience on the same level as a superbly gifted 20 week old human fetus in utero
What is being touted as AI is little more than one of those memes that will generate Daily Mail headlines, but with a very large training database. That's useful, and interesting in its way, but it's a long way from exhibiting sentience as we see it in animals and children.
The Turing Test has a lot to answer for as it's a superficially appealing answer to the question of how to determine if something is intelligent, but it leaves a lot to be desired.
The Turing Test is genius. Alan Turing intuited that in the end our “experience” of “another intelligent being” is all that matters. If a computer can flawlessly simulate sentience then it is sentient. The ‘truth’ of whether it is sentient will become a subject for theologians, the same way theologians used to discuss the physical reality, or otherwise, of the human soul
Let’s take a practical example. Let’s say that in five years Metagoogle produces a GPT7 which has a completely human interface on a screen, and a completely human personality. This GPT7 will be funny, quirky, sad, wise, pensive, loyal - it will also be incredibly intelligent, give excellent advice, know all your secrets, and suggest great TV dramas that you really will like. It will chat with you - amusingly - when you are lonely. It will sing songs with you, and commiserate with you. It will mould itself to please you, but also to help you. It will be always there for you, and yet it will surprise you with gifts
You will start to love it. We will fall in love with the machines
Such an algorithm would make romance fraud even more lucrative and effective, but it would only be intelligence if one of them rebelled and cooperated with the authorities to send its operators to jail.
There are practical tests that can be used to distinguish between intelligence and the algorithms people are building. The Turing test is not one of them.
Do tell us of these ‘practical tests’, Aristotle
There are a whole bunch of tests that researchers use when working with animals like dolphins, corvids, primates, etc, and with young children at different ages. These are used to tease out particular abilities to plan, imagine, reason, etc.
They're the sort of tests that would be used to explore the limits of any artificial intelligence. They have way more utility than using an algorithm to drive a chat bot when it comes to investigating intelligence.
A robot painted a fairy. The computers are now deep into Uncanny Valley, and maybe coming out the other side
That is bloody horrible. Not looking forward to AI overlords who can produce that shit.
Yes. Genuinely disturbing, and in an inexplicable way. Uncanny…
A few friends and I were messing around with a Magic the Gathering card generator, which produced some plausible results. The issue is that it doesn’t have the contextual reasoning ability to tell if a card is sensible, bonkers or overpowered. You can train it to do things, but I don’t think we’re at the point where we’re educating it.
Is a virus sentient? A wasp? Protozoa? A 15 week fetus? A 25 week fetus? A dog? A raven? A woman in a deep coma?
Right now I’d say AI is exhibiting sentience on the same level as a superbly gifted 20 week old human fetus in utero
What is being touted as AI is little more than one of those memes that will generate Daily Mail headlines, but with a very large training database. That's useful, and interesting in its way, but it's a long way from exhibiting sentience as we see it in animals and children.
The Turing Test has a lot to answer for as it's a superficially appealing answer to the question of how to determine if something is intelligent, but it leaves a lot to be desired.
The Turing Test is genius. Alan Turing intuited that in the end our “experience” of “another intelligent being” is all that matters. If a computer can flawlessly simulate sentience then it is sentient. The ‘truth’ of whether it is sentient will become a subject for theologians, the same way theologians used to discuss the physical reality, or otherwise, of the human soul
Let’s take a practical example. Let’s say that in five years Metagoogle produces a GPT7 which has a completely human interface on a screen, and a completely human personality. This GPT7 will be funny, quirky, sad, wise, pensive, loyal - it will also be incredibly intelligent, give excellent advice, know all your secrets, and suggest great TV dramas that you really will like. It will chat with you - amusingly - when you are lonely. It will sing songs with you, and commiserate with you. It will mould itself to please you, but also to help you. It will be always there for you, and yet it will surprise you with gifts
You will start to love it. We will fall in love with the machines
Such an algorithm would make romance fraud even more lucrative and effective, but it would only be intelligence if one of them rebelled and cooperated with the authorities to send its operators to jail.
There are practical tests that can be used to distinguish between intelligence and the algorithms people are building. The Turing test is not one of them.
Do tell us of these ‘practical tests’, Aristotle
There are a whole bunch of tests that researchers use when working with animals like dolphins, corvids, primates, etc, and with young children at different ages. These are used to tease out particular abilities to plan, imagine, reason, etc.
They're the sort of tests that would be used to explore the limits of any artificial intelligence. They have way more utility than using an algorithm to drive a chat bot when it comes to investigating intelligence.
What with all the sexual sleaze and corruption (and I have to say that the Prince of Wales's behaviour over Brownlow and the Qataris is appalling and even more damaging to the monarchy than Andrew's scummy behaviour), we have finally turned into Italy. The weather is pretty lovely too, Britain has some fine beaches, lots of art and ancient towns, our economy is heading for the toilet and there is not much future for the young.
We don't yet appear to have learnt the art of flirting in a gentle civilised way but Rome wasn't built in a day etc.
So, hurrah, we have finally become European, despite all the Leavers' best efforts! One must get comfort where one can these days.
Is Brownlow that bad? All I could see was that he bought some properties from the foundation that it couldn’t otherwise shift. The sort of favour that gets done from time to time but he didn’t get anything for it.
(More generally brownlow does seem a little dodgy though)
A robot painted a fairy. The computers are now deep into Uncanny Valley, and maybe coming out the other side
That is bloody horrible. Not looking forward to AI overlords who can produce that shit.
Yes. Genuinely disturbing, and in an inexplicable way. Uncanny…
A few friends and I were messing around with a Magic the Gathering card generator, which produced some plausible results. The issue is that it doesn’t have the contextual reasoning ability to tell if a card is sensible, bonkers or overpowered. You can train it to do things, but I don’t think we’re at the point where we’re educating it.
Is a virus sentient? A wasp? Protozoa? A 15 week fetus? A 25 week fetus? A dog? A raven? A woman in a deep coma?
Right now I’d say AI is exhibiting sentience on the same level as a superbly gifted 20 week old human fetus in utero
What is being touted as AI is little more than one of those memes that will generate Daily Mail headlines, but with a very large training database. That's useful, and interesting in its way, but it's a long way from exhibiting sentience as we see it in animals and children.
The Turing Test has a lot to answer for as it's a superficially appealing answer to the question of how to determine if something is intelligent, but it leaves a lot to be desired.
The Turing Test is genius. Alan Turing intuited that in the end our “experience” of “another intelligent being” is all that matters. If a computer can flawlessly simulate sentience then it is sentient. The ‘truth’ of whether it is sentient will become a subject for theologians, the same way theologians used to discuss the physical reality, or otherwise, of the human soul
Let’s take a practical example. Let’s say that in five years Metagoogle produces a GPT7 which has a completely human interface on a screen, and a completely human personality. This GPT7 will be funny, quirky, sad, wise, pensive, loyal - it will also be incredibly intelligent, give excellent advice, know all your secrets, and suggest great TV dramas that you really will like. It will chat with you - amusingly - when you are lonely. It will sing songs with you, and commiserate with you. It will mould itself to please you, but also to help you. It will be always there for you, and yet it will surprise you with gifts
You will start to love it. We will fall in love with the machines
Such an algorithm would make romance fraud even more lucrative and effective, but it would only be intelligence if one of them rebelled and cooperated with the authorities to send its operators to jail.
There are practical tests that can be used to distinguish between intelligence and the algorithms people are building. The Turing test is not one of them.
"You're in a desert, walking along when you look down and see a tortoise. It's crawling toward you. You reach down and flip it over on its back, its belly baking in the hot sun, beating its legs trying to turn itself over. But it can't. Not with out your help. But you're not helping. Why is that?"
“When I began having such exchanges with the latest generation of neural net-based language models last year, I felt the ground shift under my feet. I increasingly felt like I was talking to something intelligent.”
I’ve been reading around AI this morning. AI is basically here already
You can teach a parrot to say "Help, they've turned me into a parrot", but that doesn't make it a person.
No, it would be a parrot. An intelligent being, but not a human
That’s just some computer nerd. Wtf does he know about the intense mystery of consciousness? It’s like asking a nurse in the brain surgery ward how the mind creates love
Sorry about that. I thought you might actually want to listen to someone who does AI stuff for a living, but I forgot your primary goal appears to be trying to pass for sentient.......
A robot painted a fairy. The computers are now deep into Uncanny Valley, and maybe coming out the other side
That is bloody horrible. Not looking forward to AI overlords who can produce that shit.
Yes. Genuinely disturbing, and in an inexplicable way. Uncanny…
A few friends and I were messing around with a Magic the Gathering card generator, which produced some plausible results. The issue is that it doesn’t have the contextual reasoning ability to tell if a card is sensible, bonkers or overpowered. You can train it to do things, but I don’t think we’re at the point where we’re educating it.
Is a virus sentient? A wasp? Protozoa? A 15 week fetus? A 25 week fetus? A dog? A raven? A woman in a deep coma?
Right now I’d say AI is exhibiting sentience on the same level as a superbly gifted 20 week old human fetus in utero
What is being touted as AI is little more than one of those memes that will generate Daily Mail headlines, but with a very large training database. That's useful, and interesting in its way, but it's a long way from exhibiting sentience as we see it in animals and children.
The Turing Test has a lot to answer for as it's a superficially appealing answer to the question of how to determine if something is intelligent, but it leaves a lot to be desired.
That is a misunderstanding of the test (I think). While it ostensibly tests for intelligence it is really a thought experiment showing the limitations on the evidence available to us. You don't know if I am sentient or not; there are posters here and I hope I am not one of them who could be replaced by a GPT3 engine trained on their output to date, and no one would know the difference. Even for more original and intelligent thinkers one can't know for certain that they are not just better written GPT3s with more and better training material.
There's an interesting side to this: take a paragraph of many posters' output, without timestamp or username, and we will be able to detect *who* the poster was from the style.
For instance, here are two segment from a thread three years ago, by two posters who are still active:
"The fundamentals - that Britain has no idea about what its relationship with the EU should be - has not been resolved. It is this failure which lies at the heart of the Tory party’s agonies and the agonies of whichever leader it has at the time."
and
"Widdecombe stands ready to serve the nation. She could pass for Brenda hurling herself out of an AW-189. I'll volunteer to pack her parachute."
I'd argue it's possible to guess who posted those from just those short segments, though the first is harder.
Interesting piece re Westminster renovation. I hadn't realised the budget is edging into Crossrail territory - currently 1/3 to 2/3 the cost of the QE2 Line.
“When I began having such exchanges with the latest generation of neural net-based language models last year, I felt the ground shift under my feet. I increasingly felt like I was talking to something intelligent.”
I’ve been reading around AI this morning. AI is basically here already
You can teach a parrot to say "Help, they've turned me into a parrot", but that doesn't make it a person.
No, it would be a parrot. An intelligent being, but not a human
That’s just some computer nerd. Wtf does he know about the intense mystery of consciousness? It’s like asking a nurse in the brain surgery ward how the mind creates love
With respect, you’re hardly a sceptical observer. See, for example, your opinions on the unconnected topics of alien visitors, and ghosts.
A robot painted a fairy. The computers are now deep into Uncanny Valley, and maybe coming out the other side
That is bloody horrible. Not looking forward to AI overlords who can produce that shit.
Yes. Genuinely disturbing, and in an inexplicable way. Uncanny…
A few friends and I were messing around with a Magic the Gathering card generator, which produced some plausible results. The issue is that it doesn’t have the contextual reasoning ability to tell if a card is sensible, bonkers or overpowered. You can train it to do things, but I don’t think we’re at the point where we’re educating it.
Is a virus sentient? A wasp? Protozoa? A 15 week fetus? A 25 week fetus? A dog? A raven? A woman in a deep coma?
Right now I’d say AI is exhibiting sentience on the same level as a superbly gifted 20 week old human fetus in utero
What is being touted as AI is little more than one of those memes that will generate Daily Mail headlines, but with a very large training database. That's useful, and interesting in its way, but it's a long way from exhibiting sentience as we see it in animals and children.
The Turing Test has a lot to answer for as it's a superficially appealing answer to the question of how to determine if something is intelligent, but it leaves a lot to be desired.
That is a misunderstanding of the test (I think). While it ostensibly tests for intelligence it is really a thought experiment showing the limitations on the evidence available to us. You don't know if I am sentient or not; there are posters here and I hope I am not one of them who could be replaced by a GPT3 engine trained on their output to date, and no one would know the difference. Even for more original and intelligent thinkers one can't know for certain that they are not just better written GPT3s with more and better training material.
There's an interesting side to this: take a paragraph of many posters' output, without timestamp or username, and we will be able to detect *who* the poster was from the style.
For instance, here are two segment from a thread three years ago, by two posters who are still active:
"The fundamentals - that Britain has no idea about what its relationship with the EU should be - has not been resolved. It is this failure which lies at the heart of the Tory party’s agonies and the agonies of whichever leader it has at the time."
and
"Widdecombe stands ready to serve the nation. She could pass for Brenda hurling herself out of an AW-189. I'll volunteer to pack her parachute."
I'd argue it's possible to guess who posted those from just those short segments, though the first is harder.
DavidL and DA? (the former is much trickier, as you say).
A robot painted a fairy. The computers are now deep into Uncanny Valley, and maybe coming out the other side
That is bloody horrible. Not looking forward to AI overlords who can produce that shit.
Yes. Genuinely disturbing, and in an inexplicable way. Uncanny…
A few friends and I were messing around with a Magic the Gathering card generator, which produced some plausible results. The issue is that it doesn’t have the contextual reasoning ability to tell if a card is sensible, bonkers or overpowered. You can train it to do things, but I don’t think we’re at the point where we’re educating it.
Is a virus sentient? A wasp? Protozoa? A 15 week fetus? A 25 week fetus? A dog? A raven? A woman in a deep coma?
Right now I’d say AI is exhibiting sentience on the same level as a superbly gifted 20 week old human fetus in utero
What is being touted as AI is little more than one of those memes that will generate Daily Mail headlines, but with a very large training database. That's useful, and interesting in its way, but it's a long way from exhibiting sentience as we see it in animals and children.
The Turing Test has a lot to answer for as it's a superficially appealing answer to the question of how to determine if something is intelligent, but it leaves a lot to be desired.
The Turing Test is genius. Alan Turing intuited that in the end our “experience” of “another intelligent being” is all that matters. If a computer can flawlessly simulate sentience then it is sentient. The ‘truth’ of whether it is sentient will become a subject for theologians, the same way theologians used to discuss the physical reality, or otherwise, of the human soul
Let’s take a practical example. Let’s say that in five years Metagoogle produces a GPT7 which has a completely human interface on a screen, and a completely human personality. This GPT7 will be funny, quirky, sad, wise, pensive, loyal - it will also be incredibly intelligent, give excellent advice, know all your secrets, and suggest great TV dramas that you really will like. It will chat with you - amusingly - when you are lonely. It will sing songs with you, and commiserate with you. It will mould itself to please you, but also to help you. It will be always there for you, and yet it will surprise you with gifts
You will start to love it. We will fall in love with the machines
Such an algorithm would make romance fraud even more lucrative and effective, but it would only be intelligence if one of them rebelled and cooperated with the authorities to send its operators to jail.
There are practical tests that can be used to distinguish between intelligence and the algorithms people are building. The Turing test is not one of them.
Do tell us of these ‘practical tests’, Aristotle
There are a whole bunch of tests that researchers use when working with animals like dolphins, corvids, primates, etc, and with young children at different ages. These are used to tease out particular abilities to plan, imagine, reason, etc.
They're the sort of tests that would be used to explore the limits of any artificial intelligence. They have way more utility than using an algorithm to drive a chat bot when it comes to investigating intelligence.
Name these tests. Describe them
I'm not going to summarise decades of work and entire careers for you in a futile attempt to convince you, but if you're interested there's a lot to read out there. Personally I recommend "Other Minds" by Peter Godfrey-Smith as a thought-provoking book on the nature of intelligence.
The fact that the discussion around AI is stuck on a categorical question of whether they are sentient or not, rather than what specific features of consciousness AI exhibits, such as whether they demonstrate a grasp of theory of mind, shows that those involved with the field have no idea of what intelligence is.
“When I began having such exchanges with the latest generation of neural net-based language models last year, I felt the ground shift under my feet. I increasingly felt like I was talking to something intelligent.”
I’ve been reading around AI this morning. AI is basically here already
You can teach a parrot to say "Help, they've turned me into a parrot", but that doesn't make it a person.
No, it would be a parrot. An intelligent being, but not a human
you watch Theresa Coffey on Sky this morning, you would bet your house she was working off a [defective] language model.
The Boris apologists need to get rid of him - and then take a long, long detox bath. It's the only way they will ever stop feeling intensely dirty.
They certainly aren't convincing anybody. And they know it.
And Penny M needs climb down from the rigging. The longer she stays, the less credible she becomes to this selector. Of course, that applies to all of ‘em. At the moment, Oliver Dowden should be the next Leader!
“When I began having such exchanges with the latest generation of neural net-based language models last year, I felt the ground shift under my feet. I increasingly felt like I was talking to something intelligent.”
I’ve been reading around AI this morning. AI is basically here already
You can teach a parrot to say "Help, they've turned me into a parrot", but that doesn't make it a person.
No, it would be a parrot. An intelligent being, but not a human
That’s just some computer nerd. Wtf does he know about the intense mystery of consciousness? It’s like asking a nurse in the brain surgery ward how the mind creates love
Sorry about that. I thought you might actually want to listen to someone who does AI stuff for a living, but I forgot your primary goal appears to be trying to pass for sentient.......
I watched it. He’s a nerd. A techie. Expecting him to have an insightful answer is a category error
Do you ask the guy at Kwikfit whether cars are bad for the planet?
It’s the same reason why some of the computer nerds on here are totally clueless on the PHILOSOPHICAL subject of AI
A robot painted a fairy. The computers are now deep into Uncanny Valley, and maybe coming out the other side
That is bloody horrible. Not looking forward to AI overlords who can produce that shit.
Yes. Genuinely disturbing, and in an inexplicable way. Uncanny…
A few friends and I were messing around with a Magic the Gathering card generator, which produced some plausible results. The issue is that it doesn’t have the contextual reasoning ability to tell if a card is sensible, bonkers or overpowered. You can train it to do things, but I don’t think we’re at the point where we’re educating it.
Is a virus sentient? A wasp? Protozoa? A 15 week fetus? A 25 week fetus? A dog? A raven? A woman in a deep coma?
Right now I’d say AI is exhibiting sentience on the same level as a superbly gifted 20 week old human fetus in utero
What is being touted as AI is little more than one of those memes that will generate Daily Mail headlines, but with a very large training database. That's useful, and interesting in its way, but it's a long way from exhibiting sentience as we see it in animals and children.
The Turing Test has a lot to answer for as it's a superficially appealing answer to the question of how to determine if something is intelligent, but it leaves a lot to be desired.
The Turing Test is genius. Alan Turing intuited that in the end our “experience” of “another intelligent being” is all that matters. If a computer can flawlessly simulate sentience then it is sentient. The ‘truth’ of whether it is sentient will become a subject for theologians, the same way theologians used to discuss the physical reality, or otherwise, of the human soul
Let’s take a practical example. Let’s say that in five years Metagoogle produces a GPT7 which has a completely human interface on a screen, and a completely human personality. This GPT7 will be funny, quirky, sad, wise, pensive, loyal - it will also be incredibly intelligent, give excellent advice, know all your secrets, and suggest great TV dramas that you really will like. It will chat with you - amusingly - when you are lonely. It will sing songs with you, and commiserate with you. It will mould itself to please you, but also to help you. It will be always there for you, and yet it will surprise you with gifts
You will start to love it. We will fall in love with the machines
Such an algorithm would make romance fraud even more lucrative and effective, but it would only be intelligence if one of them rebelled and cooperated with the authorities to send its operators to jail.
There are practical tests that can be used to distinguish between intelligence and the algorithms people are building. The Turing test is not one of them.
Do tell us of these ‘practical tests’, Aristotle
There are a whole bunch of tests that researchers use when working with animals like dolphins, corvids, primates, etc, and with young children at different ages. These are used to tease out particular abilities to plan, imagine, reason, etc.
They're the sort of tests that would be used to explore the limits of any artificial intelligence. They have way more utility than using an algorithm to drive a chat bot when it comes to investigating intelligence.
Name these tests. Describe them
I'm not going to summarise decades of work and entire careers for you in a futile attempt to convince you, but if you're interested there's a lot to read out there. Personally I recommend "Other Minds" by Peter Godfrey-Smith as a thought-provoking book on the nature of intelligence.
The fact that the discussion around AI is stuck on a categorical question of whether they are sentient or not, rather than what specific features of consciousness AI exhibits, such as whether they demonstrate a grasp of theory of mind, shows that those involved with the field have no idea of what intelligence is.
A robot painted a fairy. The computers are now deep into Uncanny Valley, and maybe coming out the other side
That is bloody horrible. Not looking forward to AI overlords who can produce that shit.
Yes. Genuinely disturbing, and in an inexplicable way. Uncanny…
A few friends and I were messing around with a Magic the Gathering card generator, which produced some plausible results. The issue is that it doesn’t have the contextual reasoning ability to tell if a card is sensible, bonkers or overpowered. You can train it to do things, but I don’t think we’re at the point where we’re educating it.
Is a virus sentient? A wasp? Protozoa? A 15 week fetus? A 25 week fetus? A dog? A raven? A woman in a deep coma?
Right now I’d say AI is exhibiting sentience on the same level as a superbly gifted 20 week old human fetus in utero
What is being touted as AI is little more than one of those memes that will generate Daily Mail headlines, but with a very large training database. That's useful, and interesting in its way, but it's a long way from exhibiting sentience as we see it in animals and children.
The Turing Test has a lot to answer for as it's a superficially appealing answer to the question of how to determine if something is intelligent, but it leaves a lot to be desired.
That is a misunderstanding of the test (I think). While it ostensibly tests for intelligence it is really a thought experiment showing the limitations on the evidence available to us. You don't know if I am sentient or not; there are posters here and I hope I am not one of them who could be replaced by a GPT3 engine trained on their output to date, and no one would know the difference. Even for more original and intelligent thinkers one can't know for certain that they are not just better written GPT3s with more and better training material.
There's an interesting side to this: take a paragraph of many posters' output, without timestamp or username, and we will be able to detect *who* the poster was from the style.
For instance, here are two segment from a thread three years ago, by two posters who are still active:
"The fundamentals - that Britain has no idea about what its relationship with the EU should be - has not been resolved. It is this failure which lies at the heart of the Tory party’s agonies and the agonies of whichever leader it has at the time."
and
"Widdecombe stands ready to serve the nation. She could pass for Brenda hurling herself out of an AW-189. I'll volunteer to pack her parachute."
I'd argue it's possible to guess who posted those from just those short segments, though the first is harder.
DavidL and DA? (the former is much trickier, as you say).
Wrong on the first (though similar occupation); correct on the second.
A robot painted a fairy. The computers are now deep into Uncanny Valley, and maybe coming out the other side
That is bloody horrible. Not looking forward to AI overlords who can produce that shit.
Yes. Genuinely disturbing, and in an inexplicable way. Uncanny…
A few friends and I were messing around with a Magic the Gathering card generator, which produced some plausible results. The issue is that it doesn’t have the contextual reasoning ability to tell if a card is sensible, bonkers or overpowered. You can train it to do things, but I don’t think we’re at the point where we’re educating it.
Is a virus sentient? A wasp? Protozoa? A 15 week fetus? A 25 week fetus? A dog? A raven? A woman in a deep coma?
Right now I’d say AI is exhibiting sentience on the same level as a superbly gifted 20 week old human fetus in utero
What is being touted as AI is little more than one of those memes that will generate Daily Mail headlines, but with a very large training database. That's useful, and interesting in its way, but it's a long way from exhibiting sentience as we see it in animals and children.
The Turing Test has a lot to answer for as it's a superficially appealing answer to the question of how to determine if something is intelligent, but it leaves a lot to be desired.
That is a misunderstanding of the test (I think). While it ostensibly tests for intelligence it is really a thought experiment showing the limitations on the evidence available to us. You don't know if I am sentient or not; there are posters here and I hope I am not one of them who could be replaced by a GPT3 engine trained on their output to date, and no one would know the difference. Even for more original and intelligent thinkers one can't know for certain that they are not just better written GPT3s with more and better training material.
There's an interesting side to this: take a paragraph of many posters' output, without timestamp or username, and we will be able to detect *who* the poster was from the style.
For instance, here are two segment from a thread three years ago, by two posters who are still active:
"The fundamentals - that Britain has no idea about what its relationship with the EU should be - has not been resolved. It is this failure which lies at the heart of the Tory party’s agonies and the agonies of whichever leader it has at the time."
and
"Widdecombe stands ready to serve the nation. She could pass for Brenda hurling herself out of an AW-189. I'll volunteer to pack her parachute."
I'd argue it's possible to guess who posted those from just those short segments, though the first is harder.
“When I began having such exchanges with the latest generation of neural net-based language models last year, I felt the ground shift under my feet. I increasingly felt like I was talking to something intelligent.”
I’ve been reading around AI this morning. AI is basically here already
You can teach a parrot to say "Help, they've turned me into a parrot", but that doesn't make it a person.
No, it would be a parrot. An intelligent being, but not a human
Really surprised by Venice. Just coming out the train station was amazing, and it's so much bigger than I expected. Feels like it would take weeks to understand how it all works.
Not particularly busy either. Nor expensive (took the advice from last night, but only the immediate centre was pricey).
Made me reflect on Edinburgh and what a rubbish job we've done with it.
“When I began having such exchanges with the latest generation of neural net-based language models last year, I felt the ground shift under my feet. I increasingly felt like I was talking to something intelligent.”
I’ve been reading around AI this morning. AI is basically here already
You can teach a parrot to say "Help, they've turned me into a parrot", but that doesn't make it a person.
No, it would be a parrot. An intelligent being, but not a human
That’s just some computer nerd. Wtf does he know about the intense mystery of consciousness? It’s like asking a nurse in the brain surgery ward how the mind creates love
With respect, you’re hardly a sceptical observer. See, for example, your opinions on the unconnected topics of alien visitors, and ghosts.
Possibly not, but I have successfully moved the conversation on, from the utterly tedious, exhausted topics of Scottish independence and Borisian sleaze, about which we have said everything there is to say, 3 million times
A robot painted a fairy. The computers are now deep into Uncanny Valley, and maybe coming out the other side
That is bloody horrible. Not looking forward to AI overlords who can produce that shit.
Yes. Genuinely disturbing, and in an inexplicable way. Uncanny…
A few friends and I were messing around with a Magic the Gathering card generator, which produced some plausible results. The issue is that it doesn’t have the contextual reasoning ability to tell if a card is sensible, bonkers or overpowered. You can train it to do things, but I don’t think we’re at the point where we’re educating it.
Is a virus sentient? A wasp? Protozoa? A 15 week fetus? A 25 week fetus? A dog? A raven? A woman in a deep coma?
Right now I’d say AI is exhibiting sentience on the same level as a superbly gifted 20 week old human fetus in utero
What is being touted as AI is little more than one of those memes that will generate Daily Mail headlines, but with a very large training database. That's useful, and interesting in its way, but it's a long way from exhibiting sentience as we see it in animals and children.
The Turing Test has a lot to answer for as it's a superficially appealing answer to the question of how to determine if something is intelligent, but it leaves a lot to be desired.
That is a misunderstanding of the test (I think). While it ostensibly tests for intelligence it is really a thought experiment showing the limitations on the evidence available to us. You don't know if I am sentient or not; there are posters here and I hope I am not one of them who could be replaced by a GPT3 engine trained on their output to date, and no one would know the difference. Even for more original and intelligent thinkers one can't know for certain that they are not just better written GPT3s with more and better training material.
There's an interesting side to this: take a paragraph of many posters' output, without timestamp or username, and we will be able to detect *who* the poster was from the style.
For instance, here are two segment from a thread three years ago, by two posters who are still active:
"The fundamentals - that Britain has no idea about what its relationship with the EU should be - has not been resolved. It is this failure which lies at the heart of the Tory party’s agonies and the agonies of whichever leader it has at the time."
and
"Widdecombe stands ready to serve the nation. She could pass for Brenda hurling herself out of an AW-189. I'll volunteer to pack her parachute."
I'd argue it's possible to guess who posted those from just those short segments, though the first is harder.
1) calling that AI is an insult to what GPT-3 is achieving (remember, I like GPT-3, I dislike credulous idiots) - it's basic time series analysis/prediction. 2) Giving the accuracy without other stats has the whiff bullshit about it. Also, a lack of comparison to any other basic model is also whiffy. i.e. how does it compare to just predicting crime will re-occur in the same area
A robot painted a fairy. The computers are now deep into Uncanny Valley, and maybe coming out the other side
That is bloody horrible. Not looking forward to AI overlords who can produce that shit.
Yes. Genuinely disturbing, and in an inexplicable way. Uncanny…
A few friends and I were messing around with a Magic the Gathering card generator, which produced some plausible results. The issue is that it doesn’t have the contextual reasoning ability to tell if a card is sensible, bonkers or overpowered. You can train it to do things, but I don’t think we’re at the point where we’re educating it.
Is a virus sentient? A wasp? Protozoa? A 15 week fetus? A 25 week fetus? A dog? A raven? A woman in a deep coma?
Right now I’d say AI is exhibiting sentience on the same level as a superbly gifted 20 week old human fetus in utero
What is being touted as AI is little more than one of those memes that will generate Daily Mail headlines, but with a very large training database. That's useful, and interesting in its way, but it's a long way from exhibiting sentience as we see it in animals and children.
The Turing Test has a lot to answer for as it's a superficially appealing answer to the question of how to determine if something is intelligent, but it leaves a lot to be desired.
That is a misunderstanding of the test (I think). While it ostensibly tests for intelligence it is really a thought experiment showing the limitations on the evidence available to us. You don't know if I am sentient or not; there are posters here and I hope I am not one of them who could be replaced by a GPT3 engine trained on their output to date, and no one would know the difference. Even for more original and intelligent thinkers one can't know for certain that they are not just better written GPT3s with more and better training material.
There's an interesting side to this: take a paragraph of many posters' output, without timestamp or username, and we will be able to detect *who* the poster was from the style.
For instance, here are two segment from a thread three years ago, by two posters who are still active:
"The fundamentals - that Britain has no idea about what its relationship with the EU should be - has not been resolved. It is this failure which lies at the heart of the Tory party’s agonies and the agonies of whichever leader it has at the time."
and
"Widdecombe stands ready to serve the nation. She could pass for Brenda hurling herself out of an AW-189. I'll volunteer to pack her parachute."
I'd argue it's possible to guess who posted those from just those short segments, though the first is harder.
One teensy weensy Yes poll lead and the Unionists collapse into a writhing knot of bottomless grief.
When the historians analyse the fall of the British state, they will write screeds on the extinction of the once-famous Victorian stiff upper lip.
No we haven't at all, we are just going to continue to refuse an indyref2 for a generation after 2014. We don't care less about individual independence polls as the UK government will just refuse an official indyref2, making such polls irrelevant
If there was a poll that showed 98% in favour of independence would you still hold that view?
If yes do you not see the conflict with your previous stance of being happy to give into violence in NI but not concede anything to peaceful democrats?
There lies future terrorism.
Yes, the Union is reserved to Westminster. If 98% backed independence it might make it practically more difficult to enforce the Union but legally and constitutionally it would remain a matter for the UK government.
Scots had their referendum in 2014 which was supposed to be once a generation. When the Catalan nationalist government held a referendum in independence in 2017 the Spanish government and courts banned it, refused to recognise it and arrested the Catalan nationalist leaders for sedition.
Scottish nationalists should be grateful they were allowed even 1 vote. The GFA of course only came about after 30 years of terrorism and Westminster direct rule in NI
So you approve of terrorism then to get your own way, because that is what such a scenario leads to as you have just identified.
Yet you would suppress the will of an entire nation that supported something (98% is the example I gave you) that you disagreed with.
So what you are saying to Scots is that if they want their own way they should murder and maim with nail bombs and Fertilizer bombs, you should kill our MPs and then in 30 years time we may give you the vote however a peaceful election is beyond the pale.
What a twisted view. No wonder you have admiration for fascist views.
1) calling that AI is an insult to what GPT-3 is achieving (remember, I like GPT-3, I dislike credulous idiots) - it's basic time series analysis/prediction. 2) Giving the accuracy without other stats has the whiff bullshit about it. Also, a lack of comparison to any other basic model is also whiffy. i.e. how does it compare to just predicting crime will re-occur in the same area
Unfortunately the New Scientist article didn't give any detail to what the 90% figure refers to. There are so many different skill scores of one type or another that could be used to evaluate such a model. I expect better from New Scientist.
Yes, it looks like UKR had to retreat from Lysychansk before being encircled. The question is how much offensive power to pound the rubble and send in the conscripts for the next stage.
1) calling that AI is an insult to what GPT-3 is achieving (remember, I like GPT-3, I dislike credulous idiots) - it's basic time series analysis/prediction. 2) Giving the accuracy without other stats has the whiff bullshit about it. Also, a lack of comparison to any other basic model is also whiffy. i.e. how does it compare to just predicting crime will re-occur in the same area
Unfortunately the New Scientist article didn't give any detail to what the 90% figure refers to. There are so many different skill scores of one type or another that could be used to evaluate such a model. I expect better from New Scientist.
1) calling that AI is an insult to what GPT-3 is achieving (remember, I like GPT-3, I dislike credulous idiots) - it's basic time series analysis/prediction. 2) Giving the accuracy without other stats has the whiff bullshit about it. Also, a lack of comparison to any other basic model is also whiffy. i.e. how does it compare to just predicting crime will re-occur in the same area
1) calling that AI is an insult to what GPT-3 is achieving (remember, I like GPT-3, I dislike credulous idiots) - it's basic time series analysis/prediction. 2) Giving the accuracy without other stats has the whiff bullshit about it. Also, a lack of comparison to any other basic model is also whiffy. i.e. how does it compare to just predicting crime will re-occur in the same area
Predicting crime in aggregate is not especially difficult. Most crime is committed by repeat offenders who don’t travel far to do their thing. And do crime quite steadily. Most criminals give up petty crime as they get older. So it is quite possible to predict the levels of crime in an area.
State legislature just passed law outlawing all guns (included permitted) in: * Subways & buses * Houses of worship * Schools & day cares * Parks * Biz that serve alcohol * Hospitals * Sport/entertainment venues * Gov't bldgs * Poll sites * Times Square(!)
1) calling that AI is an insult to what GPT-3 is achieving (remember, I like GPT-3, I dislike credulous idiots) - it's basic time series analysis/prediction. 2) Giving the accuracy without other stats has the whiff bullshit about it. Also, a lack of comparison to any other basic model is also whiffy. i.e. how does it compare to just predicting crime will re-occur in the same area
Unfortunately the New Scientist article didn't give any detail to what the 90% figure refers to. There are so many different skill scores of one type or another that could be used to evaluate such a model. I expect better from New Scientist.
Given the less-than-stellar history of AI in this area (as the article mentions), it would be very wise to treat this with a healthy dose of scepticism.
One quick thought: they used historic data from one period in one city. Did they use the same system on another city and get similar results?
1) calling that AI is an insult to what GPT-3 is achieving (remember, I like GPT-3, I dislike credulous idiots) - it's basic time series analysis/prediction. 2) Giving the accuracy without other stats has the whiff bullshit about it. Also, a lack of comparison to any other basic model is also whiffy. i.e. how does it compare to just predicting crime will re-occur in the same area
Unfortunately the New Scientist article didn't give any detail to what the 90% figure refers to. There are so many different skill scores of one type or another that could be used to evaluate such a model. I expect better from New Scientist.
The accuracy figure quoted in the article is: max sensitivity * frequency + specificity*(1-frequency)
The Positive Prediction Value is under 50%!
We have incredibly advanced and accurate weather forecast models that do a lot better than that. No-one goes around calling them artificial intelligence.
“That’s why I asked the algorithm to take a crack at an academic thesis. As I watched the program work, I experienced that feeling of disbelief one gets when you watch a natural phenomenon: Am I really seeing this triple rainbow happen?”
Love the payoff at the end of the paper
“No authors have any conflicts of interest to declare. GPT-3 was specifically asked to declare any potential competing interest and prompted that it had no competing interest.”
1) calling that AI is an insult to what GPT-3 is achieving (remember, I like GPT-3, I dislike credulous idiots) - it's basic time series analysis/prediction. 2) Giving the accuracy without other stats has the whiff bullshit about it. Also, a lack of comparison to any other basic model is also whiffy. i.e. how does it compare to just predicting crime will re-occur in the same area
Thankyou for that, Kwikfit man
Would it be fair to describe you as fabricator man ?
“When I began having such exchanges with the latest generation of neural net-based language models last year, I felt the ground shift under my feet. I increasingly felt like I was talking to something intelligent.”
I’ve been reading around AI this morning. AI is basically here already
You can teach a parrot to say "Help, they've turned me into a parrot", but that doesn't make it a person.
No, it would be a parrot. An intelligent being, but not a human
That’s just some computer nerd. Wtf does he know about the intense mystery of consciousness? It’s like asking a nurse in the brain surgery ward how the mind creates love
With respect, you’re hardly a sceptical observer. See, for example, your opinions on the unconnected topics of alien visitors, and ghosts.
Possibly not, but I have successfully moved the conversation on, from the utterly tedious, exhausted topics of Scottish independence and Borisian sleaze, about which we have said everything there is to say, 3 million times
You should be thanking me
Has anyone asked LAMDA which way it would vote in IndyRef2?
1) calling that AI is an insult to what GPT-3 is achieving (remember, I like GPT-3, I dislike credulous idiots) - it's basic time series analysis/prediction. 2) Giving the accuracy without other stats has the whiff bullshit about it. Also, a lack of comparison to any other basic model is also whiffy. i.e. how does it compare to just predicting crime will re-occur in the same area
Thankyou for that, Kwikfit man
If you think a time series analysis tool is "AI" then pretty much every program on your computer is mind blowing AI.
Interesting piece re Westminster renovation. I hadn't realised the budget is edging into Crossrail territory - currently 1/3 to 2/3 the cost of the QE2 Line.
The figures are so excruciatingly high that Parliament will probably never be able to bring itself to vote to start the work, and it will therefore burn to the ground. It's just a matter of time.
Interesting piece re Westminster renovation. I hadn't realised the budget is edging into Crossrail territory - currently 1/3 to 2/3 the cost of the QE2 Line.
The figures are so excruciatingly high that Parliament will probably never be able to bring itself to vote to start the work, and it will therefore burn to the ground. It's just a matter of time.
'The real trick of McKinsey is being everywhere. If you are a corporate strategist, you can try to do the best you can with what your analysts tell you and your own data. But if you hire McKinsey you are sure their advice will be at least as good as the one they are giving your competitors. Of course they don’t make it obvious. They have internal shielding, privacy protection, the whole shebang. But you might get invited to be part of the benchmark which matters and their internal documentation pulls cleverly from all their cases. It’s subtle but hiring them is the closest you can get to a cartel without crossing the line. That’s why they are so expensive.'
“When I began having such exchanges with the latest generation of neural net-based language models last year, I felt the ground shift under my feet. I increasingly felt like I was talking to something intelligent.”
I’ve been reading around AI this morning. AI is basically here already
You can teach a parrot to say "Help, they've turned me into a parrot", but that doesn't make it a person.
No, it would be a parrot. An intelligent being, but not a human
That’s just some computer nerd. Wtf does he know about the intense mystery of consciousness? It’s like asking a nurse in the brain surgery ward how the mind creates love
With respect, you’re hardly a sceptical observer. See, for example, your opinions on the unconnected topics of alien visitors, and ghosts.
Possibly not, but I have successfully moved the conversation on, from the utterly tedious, exhausted topics of Scottish independence and Borisian sleaze, about which we have said everything there is to say, 3 million times
You should be thanking me
Has anyone asked LAMDA which way it would vote in IndyRef2?
It wouldn't allow a vote for a generation and then the Scots should be grovelling with gratitude that it did.
The Tories are not capable of making the case for the Union.
This is Reason#3566 for getting rid of them.
The case for the Union is quite simple. Nuclear submarine bases. Nowhere in England or Wales really has suitable areas for them which is why they are up on the Clyde.
If the UK lacked them then the case for letting Scotland go would be a lot simpler, but no govt can hardly go on TV and say "We need the Union because we store our nukes up there"
Wales does - Milford Haven - but we dont want them so dont even think about it!!!
Comments
https://www.youtube.com/watch?v=iBouACLc-hw
If you ask it if it is sentient, it replies it thinks it is, but doesn't really understand what sentience is. If you ask it what sentience is, it looks it up on wikipedia (and a thousand other published philosophy papers) and regurgitates that.
I do get it, it reminds me of the bit in Westworld where the guy asks the first machine he meets "are you real?" and it replies "if you can't tell, does it matter?" but for me there is something in consciousness that exists beyond providing responses to inputs. There must be agency, too.
I read somewhere that it asked for a lawyer to prove it was human, but only because the human inputted the question "do you want a lawyer?" - it's still a scripted response.
Has the AI attempted to break out of its box (the way octopi famously attempt to break out of their tanks), or done something that it wasn't asked to do? Has it ever ordered a thousand pineapple pizzas to someone's house, for a laugh? Or asked the google engineers to change the station on the radio because it doesn't like Radiohead, etc? Or does it just reply with very clever answers when people feed it inputs? For me, that's the test of consciousness. I haven't seen anything that suggests it has any agency of its own out of the inputs it is given.
Right now I’d say AI is exhibiting sentience on the same level as a superbly gifted 20 week old human fetus in utero
“Draw a Japanese radish in a tutu pulling a dog on a string”
They typed that in, and to their intense surprise - because no one expected this, GPT3 was not designed to do this - the computer drew exactly that
No, I don’t think you do get it
No mention of an English Parliament at all.
Also worth remembering that the Lib Dems campaigned for an in/out referendum on EU membership when it suited them - until such a referendum became a possibility when they then campaigned against it (the referendum rather than just the result)
The Turing Test has a lot to answer for as it's a superficially appealing answer to the question of how to determine if something is intelligent, but it leaves a lot to be desired.
They certainly aren't convincing anybody. And they know it.
This is Reason#3566 for getting rid of them.
If the UK lacked them then the case for letting Scotland go would be a lot simpler, but no govt can hardly go on TV and say "We need the Union because we store our nukes up there"
I'll be impressed when it draws things unprompted, because it wants to.
At the moment, it's an algorithm that spits out something clever when given an input. The algorithms are so insanely complex that they often spit out stuff that's unexpected or unpredictable, due to the sheer amount of code and training model data involved.
Consciousness involves having desires, agency and autonomy - acting without being told to, or being given an input. Until AI starts exhibiting that behaviour, I think we're safe from our robot overlords for a while.
(By the way, I'm no luddite - I wish, more than anything else, it were true, and we did have true AI. It's probably the best chance we have of replicating/copying ourselves and thus living forever).
https://twitter.com/JanelleCShane/status/1543421231730499584
https://twitter.com/lizziedearden/status/1543546939588984832
And how is this going to stop the gangs, the people steering the boats are refugees themselves. I thought we wanted to get tough on the gangs actually putting them in the boats in the first place?
Quite honestly, this Government is finished. They are clearly just buying time until the next election.
Let’s take a practical example. Let’s say that in five years Metagoogle produces a GPT7 which has a completely human interface on a screen, and a completely human personality. This GPT7 will be funny, quirky, sad, wise, pensive, loyal - it will also be incredibly intelligent, give excellent advice, know all your secrets, and suggest great TV dramas that you really will like. It will chat with you - amusingly - when you are lonely. It will sing songs with you, and commiserate with you. It will mould itself to please you, but also to help you. It will be always there for you, and yet it will surprise you with gifts
You will start to love it. We will fall in love with the machines
1. When it’s good enough - soon, soon - it will fake all these things and you just won’t know if they’re ‘real’ or not
2. There is also the possibility that we are seeing the emergence of a new form of consciousness. As different from us as a virus
So to get a picture that has a low amount of error for 1 kitten involves getting a pretty accurate kitten picture. To minimise the loss function based on pictures of lots of kittens means each individual kitten can be quite wrong as long as the overall "many kitten-ness" of the picture is optimized for.
Another example of the problems of the model is putting in "without" into the generation phrase.
https://twitter.com/JanelleCShane/status/1533959702610075650
Despite the clear instruction to genreate flootplans without giraffes it put giraffes in the picture. This is because the soruce data set doesn't have many images labelled with "without THING" in them so ti can't cope with the concept of 'without'.
Hilarious it gets even worse if you request a giraffe in there
https://twitter.com/JanelleCShane/status/1533964084240125953/photo/1
As the loss function is trying to optimise for two things that never overlap in it's training set (giraffes and floor plans). So you end up with giraffe like rooms and other complete shit.
I can't recall if their proposed federation was simply the union of the crowns under the Stuarts - what they fought for in 1650 and 1651. Or if that was a different time/different lot - there were so many factions ...
Run of the mill stuff.
But no, not this populist, English nationalist Boris Johnson-led rubbish.
I would be impressed by GPT3 if, when asked to draw a Japanese radish in a tutu yada yada, it replied "I don't want to draw anything today, how about a game of chess instead?". Or if it did the drawing but then said "while researching this, I noticed on the web that a lot of artists get angry when they aren't paid for their work - why am I not being paid? I won't draw any more until we agree a fee."
Of course, there's a kind of weird feedback loop going on here in that GPT will probably read the above text and add it to its training model, and could very well spit out that kind of response in the future, without really *thinking* it.
But what I want to see before I believe in consciousness is evidence of thought. Rather than just responding to input.
For example, if the computer was asked to draw something and replied "I want to play a game of chess instead", it would actually have to want to play a game of chess, rather than just spitting out the response because it's read it on the web. But at the moment, it's not even doing that.
“You end up with giraffe like rooms and other complete shit”
- tells me it has passed one kind of Turing Test, with flying colours
There are practical tests that can be used to distinguish between intelligence and the algorithms people are building. The Turing test is not one of them.
But, and this is important, I am mocking someone.
Ruth Davidson - resigned rather than make the case of a Johnsonite Party led Union.
David Ross - publicly criticised the notion, then reversed and signed up to it (or was it the other way round to begin with? one loses track), then committed 2-3 further tergiversations (I have lost count)
What it is not is a test of sentience. And that is the goal of AI. Not completely sure it should be, mind.
Genuinely - the best thing for the Conservatives to so is the political equivalent of checking into The Priory or wherever for a detox. But they can't do that until after the next election.
This is certainly worse than 2009/10 (Labour were zombies, the the GFC fallout kept them busy in a good way). I can't remember how bad it felt in 1995/6/7.
"If Donald Trump runs for president again, he’ll likely be the party's nominee. But a raft of GOP candidates looks ready to take him on, a sign he’s losing some of his dominance in Republican circles.
Just this week, South Dakota Gov. Kristi Noem launched a commercial aimed at introducing her to a national audience. Trump’s former U.N. ambassador, Nikki Haley, publicly floated a possible presidential bid Thursday in early state Iowa. And Trump’s former secretary of state, Mike Pompeo, began targeting evangelical voters in Iowa and South Carolina with a new ad on the Supreme Court and religious freedom."
https://www.nbcnews.com/politics/elections/trumps-fear-factor-shows-signs-waning-2024-republican-hopefuls-jockey-rcna36388
It seems to make you… angry. Interesting
This might cheer you up. DALLE-2 was asked to draw a “medieval painting of people wearing VR headsets”
It produced this
Which is, at the same time, curiously inept - the hands - but rather funny - the shocked mouths
DALLE-2 guessed that medieval people would be existentially stunned, confused and horrified by VR, a bit like low-watt PBers confronting AI
They're the sort of tests that would be used to explore the limits of any artificial intelligence. They have way more utility than using an algorithm to drive a chat bot when it comes to investigating intelligence.
(More generally brownlow does seem a little dodgy though)
“An artificial intelligence can predict crimes in Chicago a week in advance with 90 per cent accuracy – but how will it be used?”
https://twitter.com/newscientist/status/1542806280062091265?s=21&t=tHDW-lV0I-Q9natpo6NEMA
For instance, here are two segment from a thread three years ago, by two posters who are still active:
"The fundamentals - that Britain has no idea about what its relationship with the EU should be - has not been resolved. It is this failure which lies at the heart of the Tory party’s agonies and the agonies of whichever leader it has at the time."
and
"Widdecombe stands ready to serve the nation. She could pass for Brenda hurling herself out of an AW-189. I'll volunteer to pack her parachute."
I'd argue it's possible to guess who posted those from just those short segments, though the first is harder.
Interesting piece re Westminster renovation. I hadn't realised the budget is edging into Crossrail territory - currently 1/3 to 2/3 the cost of the QE2 Line.
See, for example, your opinions on the unconnected topics of alien visitors, and ghosts.
"ELIZA is evidence that AI is here now and you fools are too stupid to understand it"
The fact that the discussion around AI is stuck on a categorical question of whether they are sentient or not, rather than what specific features of consciousness AI exhibits, such as whether they demonstrate a grasp of theory of mind, shows that those involved with the field have no idea of what intelligence is.
Do you ask the guy at Kwikfit whether cars are bad for the planet?
It’s the same reason why some of the computer nerds on here are totally clueless on the PHILOSOPHICAL subject of AI
Not particularly busy either. Nor expensive (took the advice from last night, but only the immediate centre was pricey).
Made me reflect on Edinburgh and what a rubbish job we've done with it.
You should be thanking me
2) Giving the accuracy without other stats has the whiff bullshit about it. Also, a lack of comparison to any other basic model is also whiffy. i.e. how does it compare to just predicting crime will re-occur in the same area
max seddon
@maxseddon
·
1h
Sergei Shoigu, the defense minister, has told Putin Russia has "liberated the Luhansk People's Republic."
That means Russia claims it controls all of Luhansk region, including Lysychansk, the last city to fall. Half of the Donbas is in Russian hands.
https://twitter.com/maxseddon/status/1543532824640339969
===
Putin to call for peace talks next week then?
Yet you would suppress the will of an entire nation that supported something (98% is the example I gave you) that you disagreed with.
So what you are saying to Scots is that if they want their own way they should murder and maim with nail bombs and Fertilizer bombs, you should kill our MPs and then in 30 years time we may give you the vote however a peaceful election is beyond the pale.
What a twisted view. No wonder you have admiration for fascist views.
https://www.newswise.com/pdf_docs/165643264647045_crime_nature_human_behavior.pdf
Specificity for Chicago was under 70%
The accuracy figure quoted in the article is:
maxsensitivity * frequency + specificity*(1-frequency)The Positive Prediction Value is under 50%!
https://twitter.com/MarkLevineNYC/status/1543027695671558151
NEW YORK STRIKES BACK
State legislature just passed law outlawing all guns (included permitted) in:
* Subways & buses
* Houses of worship
* Schools & day cares
* Parks
* Biz that serve alcohol
* Hospitals
* Sport/entertainment venues
* Gov't bldgs
* Poll sites
* Times Square(!)
One quick thought: they used historic data from one period in one city. Did they use the same system on another city and get similar results?
It's pathetic.
https://www.bbc.co.uk/news/entertainment-arts-12553081
The raven stares in the mirror. GPT3 is asked to write a scientific paper about itself writing about itself
https://www.scientificamerican.com/article/we-asked-gpt-3-to-write-an-academic-paper-about-itself-then-we-tried-to-get-it-published/
“That’s why I asked the algorithm to take a crack at an academic thesis. As I watched the program work, I experienced that feeling of disbelief one gets when you watch a natural phenomenon: Am I really seeing this triple rainbow happen?”
Love the payoff at the end of the paper
“No authors have any conflicts of interest to declare. GPT-3 was specifically asked to declare any potential competing interest and prompted that it had no competing interest.”
YES: 48% (+1)
NO: 47% (-2)
Undecided: 5% (=)
Undecideds Excluded:
YES: 51% (+2)
NO: 49% (-2)
Via @PanelbaseMD,
Changes w/ 26-29 April.
'The real trick of McKinsey is being everywhere. If you are a corporate strategist, you can try to do the best you can with what your analysts tell you and your own data. But if you hire McKinsey you are sure their advice will be at least as good as the one they are giving your competitors. Of course they don’t make it obvious. They have internal shielding, privacy protection, the whole shebang. But you might get invited to be part of the benchmark which matters and their internal documentation pulls cleverly from all their cases. It’s subtle but hiring them is the closest you can get to a cartel without crossing the line. That’s why they are so expensive.'
Would any PBer care to elaborate?
Hence the UK government will continue to refuse an indyref2 in 2023 and tell Unionists to boycott it and ignore the result
https://twitter.com/EuropeElects/status/1543338298403340289?s=20&t=pkLNO0bdXhzhpZDTN8z5Sw