Meta's chief AI wallah tells Musk he is wrong. Sorry @Leon
Yann LeCun
@ylecun · 4h No. If it were the case, we would have AI systems that could teach themselves to drive a car in 20 hours of practice, like any 17 year-old.
But we still don't have fully autonomous, reliable self-driving, even though we (you) have millions of hours of *labeled* training data.
A 17 year old whose prefrontal cortex isn't even fully developed can learn to drive with 20 hours of training.
AI can't learn to drive with millions of hours of training.
The idea AI is smarter than humans is facetious bullshit. It might know more, but its not smarter.
What it "knows" is, of course, the entire contents of the Internet up to last week. Reminds me of old Jowett of whom it was said "I am Master of this college; what I don't know isn't knowledge."
Argentine steaks are really good. As is their Malbec
It’s another example of a country happily fulfilling glib and simplistic preconceptions, for which we recently tried to conjure a compound German noun
Cf Colombian coffee
Cf Italians all actually being well dressed Everything being big in America Swiss products being precision engineered Guinness actually tasting better in Ireland, for reasons I’ve never fathomed Finland being full of saunas
Meta's chief AI wallah tells Musk he is wrong. Sorry @Leon
Yann LeCun
@ylecun · 4h No. If it were the case, we would have AI systems that could teach themselves to drive a car in 20 hours of practice, like any 17 year-old.
But we still don't have fully autonomous, reliable self-driving, even though we (you) have millions of hours of *labeled* training data.
Meta's chief AI wallah tells Musk he is wrong. Sorry @Leon
Yann LeCun
@ylecun · 4h No. If it were the case, we would have AI systems that could teach themselves to drive a car in 20 hours of practice, like any 17 year-old.
But we still don't have fully autonomous, reliable self-driving, even though we (you) have millions of hours of *labeled* training data.
Israel should have our complete support until Hamas is completely destroyed.
If people want to offer refuge to Palestinians outside the war zone, then that is entirely reasonable, but expecting Israel not to fight a war or to fight with its hands tied behind its back is not.
Neither should have our support they are two mediaevil theocracies in the middle east duking it out again in something like round 34.....most people dont give a shit. Leave them to it
Best analogy I heard for AI today, is that its like the invention of the calculator.
Calculators can do calculations instantly and accurately that people would struggle to do mentally. When calculators arose people were saying how bad that was for all sorts of reasons etc
But calculators are a tool we can just use everyday now and take it for granted. Just like the internet is, just like electricity is, and just like AI will be going forwards too.
Smart is knowing how to use those tools, not being able to outprocess the tools in what they've been designed to do.
My calculator can do calculations I can't in my head, but I'm smarter than the calculator as I know how to use it, not the other way around. Same with AI.
Meta's chief AI wallah tells Musk he is wrong. Sorry @Leon
Yann LeCun
@ylecun · 4h No. If it were the case, we would have AI systems that could teach themselves to drive a car in 20 hours of practice, like any 17 year-old.
But we still don't have fully autonomous, reliable self-driving, even though we (you) have millions of hours of *labeled* training data.
I think he just doesn’t grasp how fast and why AI is developing but as a proud man with a remarkable CV he’s reluctant to retire and accept his redundancy
Meta's chief AI wallah tells Musk he is wrong. Sorry @Leon
Yann LeCun
@ylecun · 4h No. If it were the case, we would have AI systems that could teach themselves to drive a car in 20 hours of practice, like any 17 year-old.
But we still don't have fully autonomous, reliable self-driving, even though we (you) have millions of hours of *labeled* training data.
Best analogy I heard for AI today, is that its like the invention of the calculator.
Calculators can do calculations instantly and accurately that people would struggle to do mentally. When calculators arose people were saying how bad that was for all sorts of reasons etc
But calculators are a tool we can just use everyday now and take it for granted. Just like the internet is, just like electricity is, and just like AI will be going forwards too.
Smart is knowing how to use those tools, not being able to outprocess the tools in what they've been designed to do.
My calculator can do calculations I can't in my head, but I'm smarter than the calculator as I know how to use it, not the other way around. Same with AI.
Stop. Just stop
“AI” is really really really really really really really really not “just like the invention of the pocket calculator”
Another so called win for the Queen of Trade Deals !
Aussie beef exports to UK 17,000 tonnes.
UK exports to Australia , a big fat zero !
And the reason , the UK signed a deal that fxcked UK farmers and allowed the Aussies to draw out for years the approval process for beef imports .
As with all the trade deals signed by the over promoted waste of space it’s all about garnering positive headlines from the right wing press .
- Total UK exports to Australia amounted to £14.3 billion in the four quarters to the end of Q3 2023 (an increase of 18.5% or £2.2 billion in current prices, compared to the four quarters to the end of Q3 2022); - Total UK imports from Australia amounted to £5.2 billion in the four quarters to the end of Q3 2023 (an increase of 15.8% or £707 million in current prices, compared to the four quarters to the end of Q3 2022)
Why the fuck are we importing beef from the other side of the world? What’s wrong with our own beef, rather than a load of probable shite from Australia.
Is it vegan?
I've never tried eating vegan food, but if I was starving I suppose I could eat a vegan as a last resort.
A falsehood, given that lots of foods one eats regularly are vegan. Never eaten potatoes, or peas, or carrots?
Oh dear.
Childish with it.
Spot the difference with and without carbs. "Then I noticed my muscles looked bigger and fuller and harder, I just couldn't believe it."
Meta's chief AI wallah tells Musk he is wrong. Sorry @Leon
Yann LeCun
@ylecun · 4h No. If it were the case, we would have AI systems that could teach themselves to drive a car in 20 hours of practice, like any 17 year-old.
But we still don't have fully autonomous, reliable self-driving, even though we (you) have millions of hours of *labeled* training data.
A 17 year old whose prefrontal cortex isn't even fully developed can learn to drive with 20 hours of training.
AI can't learn to drive with millions of hours of training.
The idea AI is smarter than humans is facetious bullshit. It might know more, but its not smarter.
Current-generation AIs don't even know what year it is. Facts are not exactly their strong point. But they haven't been designed to do so.
AI won't be smart until it can do stuff it hasn't been designed to do, simply because it wants to do it, without anyone prompting it to do so.
Like any toddler can.
Er, GPT3 did exactly that. No one taught it to draw. No one realised it had that capability. Then someone asked it to draw - and it did
“AI is just an evolution of stuff we already have, it’s not some great new thing.”
“The internet is exactly like a vegetable spiralizer”
“I can tell you for a fact, AI will provide some efficiency gains and that will be the end of it.”
The only people that think chat gpt is ai are people who know nothing on the subject and read hype from companies interested in raising money. The same happened with block chain and everything had to be block chain.
Guess what block chain was overhyped bollocks so are llms but it gives no nothing idiot journalists something to froth about and they will be just as wrong now as they were about the importance of block chain or what 3 words. The reason we despise journalists is they write shit about stuff while having no understanding and getting it wrong. I trust politicians more than journalists and I have no faith whatsoever in politicians
Meta's chief AI wallah tells Musk he is wrong. Sorry @Leon
Yann LeCun
@ylecun · 4h No. If it were the case, we would have AI systems that could teach themselves to drive a car in 20 hours of practice, like any 17 year-old.
But we still don't have fully autonomous, reliable self-driving, even though we (you) have millions of hours of *labeled* training data.
A 17 year old whose prefrontal cortex isn't even fully developed can learn to drive with 20 hours of training.
AI can't learn to drive with millions of hours of training.
The idea AI is smarter than humans is facetious bullshit. It might know more, but its not smarter.
Current-generation AIs don't even know what year it is. Facts are not exactly their strong point. But they haven't been designed to do so.
AI won't be smart until it can do stuff it hasn't been designed to do, simply because it wants to do it, without anyone prompting it to do so.
Like any toddler can.
Er, GPT3 did exactly that. No one taught it to draw. No one realised it had that capability. Then someone asked it to draw - and it did
“AI is just an evolution of stuff we already have, it’s not some great new thing.”
“The internet is exactly like a vegetable spiralizer”
“I can tell you for a fact, AI will provide some efficiency gains and that will be the end of it.”
The only people that think chat gpt is ai are people who know nothing on the subject and read hype from companies interested in raising money. The same happened with block chain and everything had to be block chain.
Guess what block chain was overhyped bollocks so are llms but it gives no nothing idiot journalists something to froth about and they will be just as wrong now as they were about the importance of block chain or what 3 words. The reason we despise journalists is they write shit about stuff while having no understanding and getting it wrong. I trust politicians more than journalists and I have no faith whatsoever in politicians
There's quite a bit of overlap between the two, though.
Meta's chief AI wallah tells Musk he is wrong. Sorry @Leon
Yann LeCun
@ylecun · 4h No. If it were the case, we would have AI systems that could teach themselves to drive a car in 20 hours of practice, like any 17 year-old.
But we still don't have fully autonomous, reliable self-driving, even though we (you) have millions of hours of *labeled* training data.
A 17 year old whose prefrontal cortex isn't even fully developed can learn to drive with 20 hours of training.
AI can't learn to drive with millions of hours of training.
The idea AI is smarter than humans is facetious bullshit. It might know more, but its not smarter.
Current-generation AIs don't even know what year it is. Facts are not exactly their strong point. But they haven't been designed to do so.
AI won't be smart until it can do stuff it hasn't been designed to do, simply because it wants to do it, without anyone prompting it to do so.
Like any toddler can.
Er, GPT3 did exactly that. No one taught it to draw. No one realised it had that capability. Then someone asked it to draw - and it did
“AI is just an evolution of stuff we already have, it’s not some great new thing.”
“The internet is exactly like a vegetable spiralizer”
“I can tell you for a fact, AI will provide some efficiency gains and that will be the end of it.”
The only people that think chat gpt is ai are people who know nothing on the subject and read hype from companies interested in raising money. The same happened with block chain and everything had to be block chain.
Guess what block chain was overhyped bollocks so are llms but it gives no nothing idiot journalists something to froth about and they will be just as wrong now as they were about the importance of block chain or what 3 words. The reason we despise journalists is they write shit about stuff while having no understanding and getting it wrong. I trust politicians more than journalists and I have no faith whatsoever in politicians
Yeah but you thought your mum was a native Cornish speaker
Checking the weather in the Alps ahead of my Easter skiing trip and by next Saturday the freezing level will be nudging 4,000 metres and temperatures in the valleys over 20C. My chosen “snow sure” resort is at 1,860 metres.
Meta's chief AI wallah tells Musk he is wrong. Sorry @Leon
Yann LeCun
@ylecun · 4h No. If it were the case, we would have AI systems that could teach themselves to drive a car in 20 hours of practice, like any 17 year-old.
But we still don't have fully autonomous, reliable self-driving, even though we (you) have millions of hours of *labeled* training data.
A 17 year old whose prefrontal cortex isn't even fully developed can learn to drive with 20 hours of training.
AI can't learn to drive with millions of hours of training.
The idea AI is smarter than humans is facetious bullshit. It might know more, but its not smarter.
Current-generation AIs don't even know what year it is. Facts are not exactly their strong point. But they haven't been designed to do so.
AI won't be smart until it can do stuff it hasn't been designed to do, simply because it wants to do it, without anyone prompting it to do so.
Like any toddler can.
Er, GPT3 did exactly that. No one taught it to draw. No one realised it had that capability. Then someone asked it to draw - and it did
“AI is just an evolution of stuff we already have, it’s not some great new thing.”
“The internet is exactly like a vegetable spiralizer”
“I can tell you for a fact, AI will provide some efficiency gains and that will be the end of it.”
"... without prompting ..."
"... someone asked it to draw ..."
So no, that's not unprompted. Someone gave it the prompt and it did it.
AI 100% is just an evolution of what we already have.
I never said the internet is exactly like a vegetable spiraliser, you did, I said its a tool and it is. Also tools are the internet, calculators and vegetable spiralisers, yes. Corkscrews too.
Meta's chief AI wallah tells Musk he is wrong. Sorry @Leon
Yann LeCun
@ylecun · 4h No. If it were the case, we would have AI systems that could teach themselves to drive a car in 20 hours of practice, like any 17 year-old.
But we still don't have fully autonomous, reliable self-driving, even though we (you) have millions of hours of *labeled* training data.
A 17 year old whose prefrontal cortex isn't even fully developed can learn to drive with 20 hours of training.
AI can't learn to drive with millions of hours of training.
The idea AI is smarter than humans is facetious bullshit. It might know more, but its not smarter.
Current-generation AIs don't even know what year it is. Facts are not exactly their strong point. But they haven't been designed to do so.
AI won't be smart until it can do stuff it hasn't been designed to do, simply because it wants to do it, without anyone prompting it to do so.
Like any toddler can.
Er, GPT3 did exactly that. No one taught it to draw. No one realised it had that capability. Then someone asked it to draw - and it did
“AI is just an evolution of stuff we already have, it’s not some great new thing.”
“The internet is exactly like a vegetable spiralizer”
“I can tell you for a fact, AI will provide some efficiency gains and that will be the end of it.”
The only people that think chat gpt is ai are people who know nothing on the subject and read hype from companies interested in raising money. The same happened with block chain and everything had to be block chain.
Guess what block chain was overhyped bollocks so are llms but it gives no nothing idiot journalists something to froth about and they will be just as wrong now as they were about the importance of block chain or what 3 words. The reason we despise journalists is they write shit about stuff while having no understanding and getting it wrong. I trust politicians more than journalists and I have no faith whatsoever in politicians
There's quite a bit of overlap between the two, though.
Messrs Gove, Johnson ...
The only difference is politicians say what they say knowing they are lying, journalists actually believe the bollocks they write and claim to be more intelligent than the rest of us whereas they are the modern day version of the village idiot
Best analogy I heard for AI today, is that its like the invention of the calculator.
Calculators can do calculations instantly and accurately that people would struggle to do mentally. When calculators arose people were saying how bad that was for all sorts of reasons etc
But calculators are a tool we can just use everyday now and take it for granted. Just like the internet is, just like electricity is, and just like AI will be going forwards too.
Smart is knowing how to use those tools, not being able to outprocess the tools in what they've been designed to do.
My calculator can do calculations I can't in my head, but I'm smarter than the calculator as I know how to use it, not the other way around. Same with AI.
Stop. Just stop
“AI” is really really really really really really really really not “just like the invention of the pocket calculator”
LLM "AI" is exactly like the invention of the pocket calculator.
Its a tool to do quickly what the user wants it to do.
Its not a sentient, self-actualising body that is doing what it wants to do all by itself.
Meta's chief AI wallah tells Musk he is wrong. Sorry @Leon
Yann LeCun
@ylecun · 4h No. If it were the case, we would have AI systems that could teach themselves to drive a car in 20 hours of practice, like any 17 year-old.
But we still don't have fully autonomous, reliable self-driving, even though we (you) have millions of hours of *labeled* training data.
A 17 year old whose prefrontal cortex isn't even fully developed can learn to drive with 20 hours of training.
AI can't learn to drive with millions of hours of training.
The idea AI is smarter than humans is facetious bullshit. It might know more, but its not smarter.
Current-generation AIs don't even know what year it is. Facts are not exactly their strong point. But they haven't been designed to do so.
AI won't be smart until it can do stuff it hasn't been designed to do, simply because it wants to do it, without anyone prompting it to do so.
Like any toddler can.
Er, GPT3 did exactly that. No one taught it to draw. No one realised it had that capability. Then someone asked it to draw - and it did
“AI is just an evolution of stuff we already have, it’s not some great new thing.”
“The internet is exactly like a vegetable spiralizer”
“I can tell you for a fact, AI will provide some efficiency gains and that will be the end of it.”
The only people that think chat gpt is ai are people who know nothing on the subject and read hype from companies interested in raising money. The same happened with block chain and everything had to be block chain.
Guess what block chain was overhyped bollocks so are llms but it gives no nothing idiot journalists something to froth about and they will be just as wrong now as they were about the importance of block chain or what 3 words. The reason we despise journalists is they write shit about stuff while having no understanding and getting it wrong. I trust politicians more than journalists and I have no faith whatsoever in politicians
Yeah but you thought your mum was a native Cornish speaker
My mother spoke a language of celtic origin yes, she said it was cornish. Tough I dont give a toss what you think isn't it why should I disbelieve her?
Meta's chief AI wallah tells Musk he is wrong. Sorry @Leon
Yann LeCun
@ylecun · 4h No. If it were the case, we would have AI systems that could teach themselves to drive a car in 20 hours of practice, like any 17 year-old.
But we still don't have fully autonomous, reliable self-driving, even though we (you) have millions of hours of *labeled* training data.
I think he just doesn’t grasp how fast and why AI is developing but as a proud man with a remarkable CV he’s reluctant to retire and accept his redundancy
Checking the weather in the Alps ahead of my Easter skiing trip and by next Saturday the freezing level will be nudging 4,000 metres and temperatures in the valleys over 20C. My chosen “snow sure” resort is at 1,860 metres.
My eldest (57) was a professional snowboarder and wintered in Chamonix but also went to Lake Tahoe and Whistler ( as a guest snowboarder) and not once was there a lack of snow
Despite the pile on on Sunak rejecting the 2nd May I would still suggest that Starmer's majority ( which he will get) will depend on just how well Reform perform
Best analogy I heard for AI today, is that its like the invention of the calculator.
Calculators can do calculations instantly and accurately that people would struggle to do mentally. When calculators arose people were saying how bad that was for all sorts of reasons etc
But calculators are a tool we can just use everyday now and take it for granted. Just like the internet is, just like electricity is, and just like AI will be going forwards too.
Smart is knowing how to use those tools, not being able to outprocess the tools in what they've been designed to do.
My calculator can do calculations I can't in my head, but I'm smarter than the calculator as I know how to use it, not the other way around. Same with AI.
Biggest indicator ai likely to be a damp squib is wall street hyping it. 2 years ago wall street was hyping the metaverse. Wall street of course overhyped the internet too. Anyone remember aol or yahoo.
Despite the pile on on Sunak rejecting the 2nd May I would still suggest that Starmer's majority ( which he will get) will depend on just how well Reform perform
Lee Anderson will be the Sir Jimmy Goldsmith of the 2024/5 election.
Despite the pile on on Sunak rejecting the 2nd May I would still suggest that Starmer's majority ( which he will get) will depend on just how well Reform perform
We have seen Reform consistently underperform their polling in byelections. Shortly we will see them do poorly in the Locals. I expect them to do so too at the GE.
To give it its due, Chat GPT has really excelled itself in this latest fictional debate between a right wing catastrophist travel journalist, a nihilist with a Cornish speaking mother, and a man who eats nothing but meat.
My father and I are both lifelong Republicans. He got aboard the “Trump Train” and I waved farewell from the station.
The sad thing is, I’ve loved talking politics with him all my life. Insightful, well-reasoned, a wealth of insight even if we didn’t agree on everything.
Now, it’s all slogans and empty bullshit bromides. Jokes about pronouns and other intellect-light culture war nonsense.
MAGA didn’t just make him angrier. It made him stupider, too. Perhaps its greatest accomplishment. 10:26 PM · Mar 14, 2024 · 794 Views
Despite the pile on on Sunak rejecting the 2nd May I would still suggest that Starmer's majority ( which he will get) will depend on just how well Reform perform
We have seen Reform consistently underperform their polling in byelections. Shortly we will see them do poorly in the Locals. I expect them to do so too at the GE.
By contrast I expect the Greens, who are at least 5 points behind Reform in polls, to gain at least 10 times as many council seats as them.
Despite the pile on on Sunak rejecting the 2nd May I would still suggest that Starmer's majority ( which he will get) will depend on just how well Reform perform
Lee Anderson will be the Sir Jimmy Goldsmith of the 2024/5 election.
I think it is calculated that The Referendum party added about 20 seats to Labour's 1997 majority, so that is an accurate analogy. But 20 years later Goldsmiths legacy was Brexit, so even though REFUK will damage the Tories in the short term, in the longer term they may well take it over.
To give it its due, Chat GPT has really excelled itself in this latest fictional debate between a right wing catastrophist travel journalist, a nihilist with a Cornish speaking mother, and a man who eats nothing but meat.
Those are all the same person, but with different usernames - all arguing with each other. Right?
Despite the pile on on Sunak rejecting the 2nd May I would still suggest that Starmer's majority ( which he will get) will depend on just how well Reform perform
We have seen Reform consistently underperform their polling in byelections. Shortly we will see them do poorly in the Locals. I expect them to do so too at the GE.
Despite the pile on on Sunak rejecting the 2nd May I would still suggest that Starmer's majority ( which he will get) will depend on just how well Reform perform
We have seen Reform consistently underperform their polling in byelections. Shortly we will see them do poorly in the Locals. I expect them to do so too at the GE.
By contrast I expect the Greens, who are at least 5 points behind Reform in polls, to gain at least 10 times as many council seats as them.
Indeed. Just for some background… Ref UK currently have 9. The Green Party of E&W currently have 744, while the Scottish Green Party and the NI Greens have 36 and 5 respectively, elected of course under STV. So the Greens have 87 times as many councillors.
Meta's chief AI wallah tells Musk he is wrong. Sorry @Leon
Yann LeCun
@ylecun · 4h No. If it were the case, we would have AI systems that could teach themselves to drive a car in 20 hours of practice, like any 17 year-old.
But we still don't have fully autonomous, reliable self-driving, even though we (you) have millions of hours of *labeled* training data.
A 17 year old whose prefrontal cortex isn't even fully developed can learn to drive with 20 hours of training.
AI can't learn to drive with millions of hours of training.
The idea AI is smarter than humans is facetious bullshit. It might know more, but its not smarter.
Current-generation AIs don't even know what year it is. Facts are not exactly their strong point. But they haven't been designed to do so.
AI won't be smart until it can do stuff it hasn't been designed to do, simply because it wants to do it, without anyone prompting it to do so.
Like any toddler can.
Er, GPT3 did exactly that. No one taught it to draw. No one realised it had that capability. Then someone asked it to draw - and it did
“AI is just an evolution of stuff we already have, it’s not some great new thing.”
“The internet is exactly like a vegetable spiralizer”
“I can tell you for a fact, AI will provide some efficiency gains and that will be the end of it.”
The only people that think chat gpt is ai are people who know nothing on the subject and read hype from companies interested in raising money. The same happened with block chain and everything had to be block chain.
Guess what block chain was overhyped bollocks so are llms but it gives no nothing idiot journalists something to froth about and they will be just as wrong now as they were about the importance of block chain or what 3 words. The reason we despise journalists is they write shit about stuff while having no understanding and getting it wrong. I trust politicians more than journalists and I have no faith whatsoever in politicians
I think there is a disparity between 'ChatGPT' and 'AI' going on. ChatGPT/Claude/Groq/Gemini/Mixtral/whatever are 'a thing'. The techniques, progress that's going on outside of that world is quite striking.
My only few IT bets down the years have been Linux, The Internet, 'Virtualisation' and 'Mobile'. I've been on a bit of a streak - but 'AI' is giving similar twitchy nose vibes,
To give it its due, Chat GPT has really excelled itself in this latest fictional debate between a right wing catastrophist travel journalist, a nihilist with a Cornish speaking mother, and a man who eats nothing but meat.
Those are all the same person, but with different usernames - all arguing with each other. Right?
Im on the lookout for the russian trolls. I will let you know when i find a likely suspect.
Best analogy I heard for AI today, is that its like the invention of the calculator.
Calculators can do calculations instantly and accurately that people would struggle to do mentally. When calculators arose people were saying how bad that was for all sorts of reasons etc
But calculators are a tool we can just use everyday now and take it for granted. Just like the internet is, just like electricity is, and just like AI will be going forwards too.
Smart is knowing how to use those tools, not being able to outprocess the tools in what they've been designed to do.
My calculator can do calculations I can't in my head, but I'm smarter than the calculator as I know how to use it, not the other way around. Same with AI.
AI is a tool and should be treated as such. The problems come when we stop doing that. I don't think it's like a picket calculator though. What AI is really, really good at is pattern matching huge amounts of data. Humans aren't so good at that.
Meta's chief AI wallah tells Musk he is wrong. Sorry @Leon
Yann LeCun
@ylecun · 4h No. If it were the case, we would have AI systems that could teach themselves to drive a car in 20 hours of practice, like any 17 year-old.
But we still don't have fully autonomous, reliable self-driving, even though we (you) have millions of hours of *labeled* training data.
A 17 year old whose prefrontal cortex isn't even fully developed can learn to drive with 20 hours of training.
AI can't learn to drive with millions of hours of training.
The idea AI is smarter than humans is facetious bullshit. It might know more, but its not smarter.
Current-generation AIs don't even know what year it is. Facts are not exactly their strong point. But they haven't been designed to do so.
AI won't be smart until it can do stuff it hasn't been designed to do, simply because it wants to do it, without anyone prompting it to do so.
Like any toddler can.
Er, GPT3 did exactly that. No one taught it to draw. No one realised it had that capability. Then someone asked it to draw - and it did
“AI is just an evolution of stuff we already have, it’s not some great new thing.”
“The internet is exactly like a vegetable spiralizer”
“I can tell you for a fact, AI will provide some efficiency gains and that will be the end of it.”
The only people that think chat gpt is ai are people who know nothing on the subject and read hype from companies interested in raising money. The same happened with block chain and everything had to be block chain.
Guess what block chain was overhyped bollocks so are llms but it gives no nothing idiot journalists something to froth about and they will be just as wrong now as they were about the importance of block chain or what 3 words. The reason we despise journalists is they write shit about stuff while having no understanding and getting it wrong. I trust politicians more than journalists and I have no faith whatsoever in politicians
Yeah but you thought your mum was a native Cornish speaker
My mother spoke a language of celtic origin yes, she said it was cornish. Tough I dont give a toss what you think isn't it why should I disbelieve her?
Don't know what they say in Cornwall, but in Ireland the Irish say: "Me mother - drunk or sober!"
Macron on French TV this evening : The war in 🇺🇦 is “existential…”If Russia were to win, life for the French would change. We would no longer have security in Europe. Who can seriously believe that Putin, who has respected no limits, would stop there?”
To give it its due, Chat GPT has really excelled itself in this latest fictional debate between a right wing catastrophist travel journalist, a nihilist with a Cornish speaking mother, and a man who eats nothing but meat.
I like the character development on the nihilist. I think we are supposed to instinctively dislike him but he often says things that show a different side to him, causing us to warm to him.
The lager-drinking device used on the catastrophist is rather obvious and grossly overdone, though.
Other parties with more councillors than Reform UK include Aspire and the Ashfield Independents, but they are doing better than the Liberal Party or the SDP.
Meta's chief AI wallah tells Musk he is wrong. Sorry @Leon
Yann LeCun
@ylecun · 4h No. If it were the case, we would have AI systems that could teach themselves to drive a car in 20 hours of practice, like any 17 year-old.
But we still don't have fully autonomous, reliable self-driving, even though we (you) have millions of hours of *labeled* training data.
A 17 year old whose prefrontal cortex isn't even fully developed can learn to drive with 20 hours of training.
AI can't learn to drive with millions of hours of training.
The idea AI is smarter than humans is facetious bullshit. It might know more, but its not smarter.
Current-generation AIs don't even know what year it is. Facts are not exactly their strong point. But they haven't been designed to do so.
AI won't be smart until it can do stuff it hasn't been designed to do, simply because it wants to do it, without anyone prompting it to do so.
Like any toddler can.
Er, GPT3 did exactly that. No one taught it to draw. No one realised it had that capability. Then someone asked it to draw - and it did
“AI is just an evolution of stuff we already have, it’s not some great new thing.”
“The internet is exactly like a vegetable spiralizer”
“I can tell you for a fact, AI will provide some efficiency gains and that will be the end of it.”
The only people that think chat gpt is ai are people who know nothing on the subject and read hype from companies interested in raising money. The same happened with block chain and everything had to be block chain.
Guess what block chain was overhyped bollocks so are llms but it gives no nothing idiot journalists something to froth about and they will be just as wrong now as they were about the importance of block chain or what 3 words. The reason we despise journalists is they write shit about stuff while having no understanding and getting it wrong. I trust politicians more than journalists and I have no faith whatsoever in politicians
I think there is a disparity between 'ChatGPT' and 'AI' going on. ChatGPT/Claude/Groq/Gemini/Mixtral/whatever are 'a thing'. The techniques, progress that's going on outside of that world is quite striking.
My only few IT bets down the years have been Linux, The Internet, 'Virtualisation' and 'Mobile'. I've been on a bit of a streak - but 'AI' is giving similar twitchy nose vibes,
(Full disclosure - my first paying gig was for Atari - so ymmv)
Meta's chief AI wallah tells Musk he is wrong. Sorry @Leon
Yann LeCun
@ylecun · 4h No. If it were the case, we would have AI systems that could teach themselves to drive a car in 20 hours of practice, like any 17 year-old.
But we still don't have fully autonomous, reliable self-driving, even though we (you) have millions of hours of *labeled* training data.
I think he just doesn’t grasp how fast and why AI is developing but as a proud man with a remarkable CV he’s reluctant to retire and accept his redundancy
I'd recommend you read this: .
We know he can cut, and we know he can paste, but does any reading take place in between?
@Nigel_Farage : 'Hope Not Hate's 'State of Hate Report' mentions me 56 times (more than any other individual), GB News is mentioned 'page after page', and people could use the word hate 'as a cover to shut down anyone they disagree with'.
While I think Reform will underperform in the May locals, they might pick up a seat or two in the London Assembly, as UKIP once did. Would that boost their national profile?
This is the RefUK candidate in my Assembly constituency: https://www.reformparty.uk/gla-barnet-and-camden He used to be in the Conservatives and, before that, supported the BNP! However, the BNP in question was the Bangladesh Nationalist Party.
While I think Reform will underperform in the May locals, they might pick up a seat or two in the London Assembly, as UKIP once did. Would that boost their national profile?
This is the RefUK candidate in my Assembly constituency: https://www.reformparty.uk/gla-barnet-and-camden He used to be in the Conservatives and, before that, supported the BNP! However, the BNP in question was the Bangladesh Nationalist Party.
His profile page says: “anti-social behaviors [sic] are in rocket-high”
While I think Reform will underperform in the May locals, they might pick up a seat or two in the London Assembly, as UKIP once did. Would that boost their national profile?
This is the RefUK candidate in my Assembly constituency: https://www.reformparty.uk/gla-barnet-and-camden He used to be in the Conservatives and, before that, supported the BNP! However, the BNP in question was the Bangladesh Nationalist Party.
He wants AI driven efficient policing, maybe Leon's right about the coming dystopia.
Macron on French TV this evening : The war in 🇺🇦 is “existential…”If Russia were to win, life for the French would change. We would no longer have security in Europe. Who can seriously believe that Putin, who has respected no limits, would stop there?”
Despite the pile on on Sunak rejecting the 2nd May I would still suggest that Starmer's majority ( which he will get) will depend on just how well Reform perform
We have seen Reform consistently underperform their polling in byelections. Shortly we will see them do poorly in the Locals. I expect them to do so too at the GE.
By contrast I expect the Greens, who are at least 5 points behind Reform in polls, to gain at least 10 times as many council seats as them.
I'm very confident that Reform will receive more votes at the GE than the Greens.
Jack and Jill have three children. Jack has three and sevenpence in his pocket, Jill has four and eightpence in her purse. How much pocket money can they give to each child?
I hesitate to insult fellow PBers by revealing the answer to be two and ninepence - you'll have worked that out for ourselves, no doubt. But I'd be interested to know if AI could do it as quickly.
Macron on French TV this evening : The war in 🇺🇦 is “existential…”If Russia were to win, life for the French would change. We would no longer have security in Europe. Who can seriously believe that Putin, who has respected no limits, would stop there?”
It is very difficult to judge how RefUK will do when they never stand. Frustrating. But following on from a previous comment, a good sign for the Greens. I find it hard to believe that come the election, RefUK will significantly outpoll them when there is no evidence of it on the ground.
So clearly most voters think both Hamas and Israel have committed war crimes and would leave it up to them to find a solution.
However I doubt it swings any seats, apart from maybe some of the Barnet seats back to the Tories if Labour push too hard for a ceasefire, or Bradford and Tower Hamlets in addition to Rochdale to Galloway if Labour are seen as too pro Israel
Macron on French TV this evening : The war in 🇺🇦 is “existential…”If Russia were to win, life for the French would change. We would no longer have security in Europe. Who can seriously believe that Putin, who has respected no limits, would stop there?”
To be fair, it has now been revealed that his pleading with Putin for peace, prior to the war starting, was at the request of the Ukrainians.
Considers matters.
Nah. He doesn’t deserve fair
Haven't you been reading our newest poster? It turns out that Bridgette is actually a man, something that M Macron was presumably unaware of when he married... him. So he deserves our deepest sympathies.
Macron on French TV this evening : The war in 🇺🇦 is “existential…”If Russia were to win, life for the French would change. We would no longer have security in Europe. Who can seriously believe that Putin, who has respected no limits, would stop there?”
To be fair, it has now been revealed that his pleading with Putin for peace, prior to the war starting, was at the request of the Ukrainians.
Considers matters.
Nah. He doesn’t deserve fair
Haven't you been reading our newest poster? It turns out that Bridgette is actually a man, something that M Macron was presumably unaware of when he married... him. So he deserves our deepest sympathies.
Macron on French TV this evening : The war in 🇺🇦 is “existential…”If Russia were to win, life for the French would change. We would no longer have security in Europe. Who can seriously believe that Putin, who has respected no limits, would stop there?”
To be fair, it has now been revealed that his pleading with Putin for peace, prior to the war starting, was at the request of the Ukrainians.
I don't get the criticism of Macron over Ukraine. He did not compromise French preparations nor support for Ukraine with weapons etc even though he was still talking to Putin every day. It strikes me as an example of exemplary statesmanship. Work for peace and prepare for war.
Both Johnson and Macron, for all their myriad other failings (Johnson more than Macron) were absolutely right with the way they handled the lead up to and start of the invasion.
Jack and Jill have three children. Jack has three and sevenpence in his pocket, Jill has four and eightpence in her purse. How much pocket money can they give to each child?
I hesitate to insult fellow PBers by revealing the answer to be two and ninepence - you'll have worked that out for ourselves, no doubt. But I'd be interested to know if AI could do it as quickly.
Yeah. It's bizarrely bad at that question. Needed four tries with extra prompting and hints.
Macron on French TV this evening : The war in 🇺🇦 is “existential…”If Russia were to win, life for the French would change. We would no longer have security in Europe. Who can seriously believe that Putin, who has respected no limits, would stop there?”
To be fair, it has now been revealed that his pleading with Putin for peace, prior to the war starting, was at the request of the Ukrainians.
I don't get the criticism of Macron over Ukraine. He did not compromise French preparations nor support for Ukraine with weapons etc even though he was still talking to Putin every day. It strikes me as an example of exemplary statesmanship. Work for peace and prepare for war.
Both Johnson and Macron, for all their myriad other failings (Johnson more than Macron) were absolutely right with the way they handled the lead up to and start of the invasion.
I think it’s just the natural mistrust of France here, combined with their undoubted commercial operations in Russia. France tends to be a ferocious defender of French commercial interests, and it was easy to assume that’s what was happening.
Jack and Jill have three children. Jack has three and sevenpence in his pocket, Jill has four and eightpence in her purse. How much pocket money can they give to each child?
I hesitate to insult fellow PBers by revealing the answer to be two and ninepence - you'll have worked that out for ourselves, no doubt. But I'd be interested to know if AI could do it as quickly.
None of GPT3.5, GPT4 or Gemini could do it, Gemini interpreted three and sevenpence as 10p (a bit like five and twenty blackbirds), the Chat GPTs both thought it was £3.07 so each child would get £2.38.
Adding in shillings where appropriate helped though Gemini for some reason decided to round up to the nearest shilling and GPT3.5 almost got it and then forgot how many pennies were in a shilling to come up with a very confident and wrong 2 shillings and 4 pence). GPT4 did get it completely right with an python code that I hadn't asked for and an odd moral lesson added.
Let's break it down. First, we need to add up Jack and Jill's total money to see what they're working with. Since we're dealing with shillings and pence, a quick refresher: there are 12 pence in a shilling.
Jack has 3 shillings and 7 pence, while Jill has 4 shillings and 8 pence. Let's convert everything into pence first for ease of calculation, then add them up, and finally divide the total by three (since they have three children) to figure out how much each child can receive. After that, we'll convert the amount back into shillings and pence for each child.
Let's calculate:
Jack and Jill can give each child 2 shillings and 9 pence. It's a tidy sum, considering the era of shillings and pence! Distributing their savings equally among their children shows a fair and practical approach to pocket money, ensuring each child receives a portion to spend or save as they see fit.
Jack and Jill have three children. Jack has three and sevenpence in his pocket, Jill has four and eightpence in her purse. How much pocket money can they give to each child?
I hesitate to insult fellow PBers by revealing the answer to be two and ninepence - you'll have worked that out for ourselves, no doubt. But I'd be interested to know if AI could do it as quickly.
Well you have seven (what?) and 15 pence, so the solution should be 2 (what) and five pence each with 1 (what) leftover.
If those were pounds you're talking about, then that's another 33 pence each with a penny left over.
It is very difficult to judge how RefUK will do when they never stand. Frustrating. But following on from a previous comment, a good sign for the Greens. I find it hard to believe that come the election, RefUK will significantly outpoll them when there is no evidence of it on the ground.
In general elections the Green vote is extremely susceptible to a tactical squeeze. Almost uniquely so, while UKIP/Brexit/Reform are almost the opposite - nihilistic protest votes.
Jack and Jill have three children. Jack has three and sevenpence in his pocket, Jill has four and eightpence in her purse. How much pocket money can they give to each child?
I hesitate to insult fellow PBers by revealing the answer to be two and ninepence - you'll have worked that out for ourselves, no doubt. But I'd be interested to know if AI could do it as quickly.
Well you have seven (what?) and 15 pence, so the solution should be 2 (what) and five pence each with 1 (what) leftover.
If those were pounds you're talking about, then that's another 33 pence each with a penny left over.
So I'm going £2.48 each with a penny leftover.
It was standard grammar to say 3 and 7 as shorthand for 3 shillings and seven pence. You'd expect an omniscient AI to know that.
Jack and Jill have three children. Jack has three and sevenpence in his pocket, Jill has four and eightpence in her purse. How much pocket money can they give to each child?
I hesitate to insult fellow PBers by revealing the answer to be two and ninepence - you'll have worked that out for ourselves, no doubt. But I'd be interested to know if AI could do it as quickly.
Well you have seven (what?) and 15 pence, so the solution should be 2 (what) and five pence each with 1 (what) leftover.
If those were pounds you're talking about, then that's another 33 pence each with a penny left over.
So I'm going £2.48 each with a penny leftover.
It was standard grammar to say 3 and 7 as shorthand for 3 shillings and seven pence. You'd expect an omniscient AI to know that.
Maybe it was in the eighteen hundreds, but to be fair the AI is getting trained on this centuries material nowadays, in which case pence go with pounds.
So £2.38 that they got is the right answer (I did a typo writing £2.48).
Macron on French TV this evening : The war in 🇺🇦 is “existential…”If Russia were to win, life for the French would change. We would no longer have security in Europe. Who can seriously believe that Putin, who has respected no limits, would stop there?”
To be fair, it has now been revealed that his pleading with Putin for peace, prior to the war starting, was at the request of the Ukrainians.
Considers matters.
Nah. He doesn’t deserve fair
Haven't you been reading our newest poster? It turns out that Bridgette is actually a man, something that M Macron was presumably unaware of when he married... him. So he deserves our deepest sympathies.
Jack and Jill have three children. Jack has three and sevenpence in his pocket, Jill has four and eightpence in her purse. How much pocket money can they give to each child?
I hesitate to insult fellow PBers by revealing the answer to be two and ninepence - you'll have worked that out for ourselves, no doubt. But I'd be interested to know if AI could do it as quickly.
Well you have seven (what?) and 15 pence, so the solution should be 2 (what) and five pence each with 1 (what) leftover.
If those were pounds you're talking about, then that's another 33 pence each with a penny left over.
So I'm going £2.48 each with a penny leftover.
It was standard grammar to say 3 and 7 as shorthand for 3 shillings and seven pence. You'd expect an omniscient AI to know that.
Maybe it was in the eighteen hundreds, but to be fair the AI is getting trained on this centuries material nowadays, in which case pence go with pounds.
So £2.38 that they got is the right answer (I did a typo writing £2.48).
£2.05 each plus 33p each from the extra pound.
It's the sort of problem I could have solved as a nine-year-old in 1960. Maybe AI is suffering from recency bias?
Jack and Jill have three children. Jack has three and sevenpence in his pocket, Jill has four and eightpence in her purse. How much pocket money can they give to each child?
I hesitate to insult fellow PBers by revealing the answer to be two and ninepence - you'll have worked that out for ourselves, no doubt. But I'd be interested to know if AI could do it as quickly.
Yeah. It's bizarrely bad at that question. Needed four tries with extra prompting and hints.
It reminded me of a particularly difficult test I tried of giving GPT3 some King William's Quiz questions in 2022.
Oddly for the question on "In 1922 how did the Seine cause 86 people to drown?" GPT3 came up with what seems like a completely made up story of a river paddle steamer full of Parisian daytrippers colliding with a tugboat with massive loss of life. Just now I did the same question to GPT4 to get the same sort of answer, just with more detail, but still completely wrong.
Jack and Jill have three children. Jack has three and sevenpence in his pocket, Jill has four and eightpence in her purse. How much pocket money can they give to each child?
I hesitate to insult fellow PBers by revealing the answer to be two and ninepence - you'll have worked that out for ourselves, no doubt. But I'd be interested to know if AI could do it as quickly.
Well you have seven (what?) and 15 pence, so the solution should be 2 (what) and five pence each with 1 (what) leftover.
If those were pounds you're talking about, then that's another 33 pence each with a penny left over.
So I'm going £2.48 each with a penny leftover.
It was standard grammar to say 3 and 7 as shorthand for 3 shillings and seven pence. You'd expect an omniscient AI to know that.
Maybe it was in the eighteen hundreds, but to be fair the AI is getting trained on this centuries material nowadays, in which case pence go with pounds.
So £2.38 that they got is the right answer (I did a typo writing £2.48).
£2.05 each plus 33p each from the extra pound.
It's the sort of problem I could have solved as a nine-year-old in 1960. Maybe AI is suffering from recency bias?
Is it bias to be grounded in this century, unless specified otherwise?
Pence have gone with pounds for my entire lifetime, let alone GPT's lifetime.
So my take on climate change is this: everywhere in the world is going to get hotter and hotter… apart from the UK which will, uniquely, contrive to get greyer and rainier and even more dismal
Oh well. At least it should discourage the boat people
I agree. Our winters are getting warmer but wetter and more miserable and our summers warmer but cloudier. Still one of the most dismal climates for its latitude in the world is Lima which despite being in the Tropics manages to be under constant grey 8 months of the year.
Indeed. Lima manages to be climatically way more miserable than anywhere in the UK, which is quite a feat given its location. It always annoys me when I read history books or guide books that reference Lima and don’t mention this fairly notable affliction
Why the fuck did the Spanish build their capital there? The incans very sensibly chose sunny and refreshing Cusco
Possibly the worst “place” I have ever been - in terms of climate meeting geography - is the desert north of Lima. The Sechura. It’s a dismal grey sand desert, strewn with trash, and cursed with that same cruel and depressing climate - chilly grey cloud like Glasgow but without the chirpy locals
Also shit food and a history of urgent child sacrifice
One of the worst drives of my life was in the desert north of lima stuck with a mad driver driving like a maniac on a narrow road and swerving at the last minute to avoid oncoming traffic. And yes its bleak and the locals are miserable.
I took this photo on a beach in deserty northern Peru as it seemed to summarise the whole place
However the little colonial towns do have a certain charm, under those sparkling blue skies
Very Quantum of Solace.
Imagine living on that street in the second photo. Wake up, put your head out of the door, hang yourself
Most people in the world live somewhere like that. It's Median Street, Planet Earth, 2024. Most people have never stayed in a hotel, owned a car, held a year's income in a bank account, etc. Yet suicide is more frequent in the USA and Canada than it is in Latin America. And the curious thing: that's with "Latin America" defined as Hispanophone and Lusophone America. There is a single country in South America that tops USA and Canada for self-topping: Guyana.
As someone who constantly travels the world, I can (happily) reassure you that most people do NOT live on a street similar to that street in the 2nd photo: ie on a dirt road, in a concrete shack, in a shitty town in the middle of an awful foggy desert. Northern Peru, as we have established, is unusually hideous
The global median person probably lives in a concrete apartment in a large city in India or China. Not beautiful, but not terrible
Hotels and resorts aren't the world. Two thirds of the population of India, the world's most populous country, live outside of cities and so presumably not in apartments. That photo could easily be from many places in Brazil (even 30 miles from the capital) or Uganda (ditto). China, yes, mostly apartments - only a third live outside of cities. Depends how you define median. And cities. I would say they're in a shitty small town somewhere in the third world. Smartphones being the hard drug that's available.
Well, if you think Uganda (GDP per capita $883) is anything like average for the world, you're insane.
There are 8 billion people on Planet Earth (give or take).
About 1.4bn of them are North America or Europe. They're pretty well off, on average. About 1.3bn are in China. They are dramatically richer than they were 20 years ago. And that's true of most of Asia.
Median incomes and lifestyles have increased extraordinarily in the last 20 years. I mean some places, like rural Russia, have gone backwards. But most places are dramatically richer than they were.
Okay. Uganda is well below average, so yes. Brazil isn't far below it, though.
I've been hunting for the world's median person. Listing countries by PPPpc and then going down the list until half the world's population is accounted for, a disgustingly quick and dirty method that has all sorts of things wrong with it, gives the following 7 countries within 10 of the median and with population >10 million:
In Asia...........Indonesia, Sri Lanka In America.....Brazil, Colombia, Peru In Africa........Algeria, South Africa
So my take on climate change is this: everywhere in the world is going to get hotter and hotter… apart from the UK which will, uniquely, contrive to get greyer and rainier and even more dismal
Oh well. At least it should discourage the boat people
I agree. Our winters are getting warmer but wetter and more miserable and our summers warmer but cloudier. Still one of the most dismal climates for its latitude in the world is Lima which despite being in the Tropics manages to be under constant grey 8 months of the year.
Indeed. Lima manages to be climatically way more miserable than anywhere in the UK, which is quite a feat given its location. It always annoys me when I read history books or guide books that reference Lima and don’t mention this fairly notable affliction
Why the fuck did the Spanish build their capital there? The incans very sensibly chose sunny and refreshing Cusco
Possibly the worst “place” I have ever been - in terms of climate meeting geography - is the desert north of Lima. The Sechura. It’s a dismal grey sand desert, strewn with trash, and cursed with that same cruel and depressing climate - chilly grey cloud like Glasgow but without the chirpy locals
Also shit food and a history of urgent child sacrifice
One of the worst drives of my life was in the desert north of lima stuck with a mad driver driving like a maniac on a narrow road and swerving at the last minute to avoid oncoming traffic. And yes its bleak and the locals are miserable.
I took this photo on a beach in deserty northern Peru as it seemed to summarise the whole place
However the little colonial towns do have a certain charm, under those sparkling blue skies
Very Quantum of Solace.
Imagine living on that street in the second photo. Wake up, put your head out of the door, hang yourself
Most people in the world live somewhere like that. It's Median Street, Planet Earth, 2024. Most people have never stayed in a hotel, owned a car, held a year's income in a bank account, etc. Yet suicide is more frequent in the USA and Canada than it is in Latin America. And the curious thing: that's with "Latin America" defined as Hispanophone and Lusophone America. There is a single country in South America that tops USA and Canada for self-topping: Guyana.
As someone who constantly travels the world, I can (happily) reassure you that most people do NOT live on a street similar to that street in the 2nd photo: ie on a dirt road, in a concrete shack, in a shitty town in the middle of an awful foggy desert. Northern Peru, as we have established, is unusually hideous
The global median person probably lives in a concrete apartment in a large city in India or China. Not beautiful, but not terrible
Hotels and resorts aren't the world. Two thirds of the population of India, the world's most populous country, live outside of cities and so presumably not in apartments. That photo could easily be from many places in Brazil (even 30 miles from the capital) or Uganda (ditto). China, yes, mostly apartments - only a third live outside of cities. Depends how you define median. And cities. I would say they're in a shitty small town somewhere in the third world. Smartphones being the hard drug that's available.
Well, if you think Uganda (GDP per capita $883) is anything like average for the world, you're insane.
There are 8 billion people on Planet Earth (give or take).
About 1.4bn of them are North America or Europe. They're pretty well off, on average. About 1.3bn are in China. They are dramatically richer than they were 20 years ago. And that's true of most of Asia.
Median incomes and lifestyles have increased extraordinarily in the last 20 years. I mean some places, like rural Russia, have gone backwards. But most places are dramatically richer than they were.
Okay. Uganda is well below average, so yes. Brazil isn't far below it, though.
I've been hunting for the world's median person. Listing countries by PPPpc and then going down the list until half the world's population is accounted for, a disgustingly quick and dirty method that has all sorts of things wrong with it, gives the following 7 countries within 10 of the median and with population >10 million:
In Asia...........Indonesia, Sri Lanka In America.....Brazil, Colombia, Peru In Africa........Algeria, South Africa
China is above, India below, by this Q&D metric.
The 2 of the 7 with closest to the median murder rate are:
Algeria, Sri Lanka
Our person may be in a small town in one of those. Where would you locate him or her?
Jack and Jill have three children. Jack has three and sevenpence in his pocket, Jill has four and eightpence in her purse. How much pocket money can they give to each child?
I hesitate to insult fellow PBers by revealing the answer to be two and ninepence - you'll have worked that out for ourselves, no doubt. But I'd be interested to know if AI could do it as quickly.
Well you have seven (what?) and 15 pence, so the solution should be 2 (what) and five pence each with 1 (what) leftover.
If those were pounds you're talking about, then that's another 33 pence each with a penny left over.
So I'm going £2.48 each with a penny leftover.
It was standard grammar to say 3 and 7 as shorthand for 3 shillings and seven pence. You'd expect an omniscient AI to know that.
Maybe it was in the eighteen hundreds, but to be fair the AI is getting trained on this centuries material nowadays, in which case pence go with pounds.
So £2.38 that they got is the right answer (I did a typo writing £2.48).
£2.05 each plus 33p each from the extra pound.
It's the sort of problem I could have solved as a nine-year-old in 1960. Maybe AI is suffering from recency bias?
Is it bias to be grounded in this century, unless specified otherwise?
Pence have gone with pounds for my entire lifetime, let alone GPT's lifetime.
Perhaps ask the Great Bot, how much in farthings is Mike Pence worth today as PB betting proposition?
Macron on French TV this evening : The war in 🇺🇦 is “existential…”If Russia were to win, life for the French would change. We would no longer have security in Europe. Who can seriously believe that Putin, who has respected no limits, would stop there?”
To be fair, it has now been revealed that his pleading with Putin for peace, prior to the war starting, was at the request of the Ukrainians.
Considers matters.
Nah. He doesn’t deserve fair
Haven't you been reading our newest poster? It turns out that Bridgette is actually a man, something that M Macron was presumably unaware of when he married... him. So he deserves our deepest sympathies.
I’m stuck in traffic on the 405
That deserves sympathy!
In the last two days, the traffic in LA has gone back to pre-pandemic levels. It's hell.
I think Rishi has got it wrong too to let it go on further .
I’m not convinced by this, “so called” confirmation. It wasn’t a planned announcement. He didn’t plan to say it or else he wouldn’t have dodged the question twenty times first. It seems an odd way to break the decision, pressed more than several times in a regional interview, and then slipping it out in frustration…
I’m also highly suspicious how the media then leapt on it, without thinking the same as what I’m thinking. He will be asked to clarify and confirm by whoever’s turn it is in the Westminster press pool tomorrow, so let’s see how unequivocal Sunak, and the no 10 briefing is.
We will at least know tomorrow either way, if they don’t row back on media interesting it’s not May second, then probably it’s not May 2nd. However if they do row back on it to the will he won’t he position again, that will mean it’s definitely May 2nd.
I think Rishi has got it wrong too to let it go on further .
I’m not convinced by this, “so called” confirmation. It wasn’t a planned announcement. He didn’t plan to say it or else he wouldn’t have dodged the question twenty times first. It seems an odd way to break the decision, pressed more than several times in a regional interview, and then slipping it out in frustration…
I’m also highly suspicious how the media then leapt on it, without thinking the same as what I’m thinking. He will be asked to clarify and confirm by whoever’s turn it is in the Westminster press pool tomorrow, so let’s see how unequivocal Sunak, and the briefing from The Mouth of Sauron is.
So my take on climate change is this: everywhere in the world is going to get hotter and hotter… apart from the UK which will, uniquely, contrive to get greyer and rainier and even more dismal
Oh well. At least it should discourage the boat people
I agree. Our winters are getting warmer but wetter and more miserable and our summers warmer but cloudier. Still one of the most dismal climates for its latitude in the world is Lima which despite being in the Tropics manages to be under constant grey 8 months of the year.
Indeed. Lima manages to be climatically way more miserable than anywhere in the UK, which is quite a feat given its location. It always annoys me when I read history books or guide books that reference Lima and don’t mention this fairly notable affliction
Why the fuck did the Spanish build their capital there? The incans very sensibly chose sunny and refreshing Cusco
Possibly the worst “place” I have ever been - in terms of climate meeting geography - is the desert north of Lima. The Sechura. It’s a dismal grey sand desert, strewn with trash, and cursed with that same cruel and depressing climate - chilly grey cloud like Glasgow but without the chirpy locals
Also shit food and a history of urgent child sacrifice
One of the worst drives of my life was in the desert north of lima stuck with a mad driver driving like a maniac on a narrow road and swerving at the last minute to avoid oncoming traffic. And yes its bleak and the locals are miserable.
I took this photo on a beach in deserty northern Peru as it seemed to summarise the whole place
However the little colonial towns do have a certain charm, under those sparkling blue skies
Very Quantum of Solace.
Imagine living on that street in the second photo. Wake up, put your head out of the door, hang yourself
Most people in the world live somewhere like that. It's Median Street, Planet Earth, 2024. Most people have never stayed in a hotel, owned a car, held a year's income in a bank account, etc. Yet suicide is more frequent in the USA and Canada than it is in Latin America. And the curious thing: that's with "Latin America" defined as Hispanophone and Lusophone America. There is a single country in South America that tops USA and Canada for self-topping: Guyana.
As someone who constantly travels the world, I can (happily) reassure you that most people do NOT live on a street similar to that street in the 2nd photo: ie on a dirt road, in a concrete shack, in a shitty town in the middle of an awful foggy desert. Northern Peru, as we have established, is unusually hideous
The global median person probably lives in a concrete apartment in a large city in India or China. Not beautiful, but not terrible
Hotels and resorts aren't the world. Two thirds of the population of India, the world's most populous country, live outside of cities and so presumably not in apartments. That photo could easily be from many places in Brazil (even 30 miles from the capital) or Uganda (ditto). China, yes, mostly apartments - only a third live outside of cities. Depends how you define median. And cities. I would say they're in a shitty small town somewhere in the third world. Smartphones being the hard drug that's available.
Well, if you think Uganda (GDP per capita $883) is anything like average for the world, you're insane.
There are 8 billion people on Planet Earth (give or take).
About 1.4bn of them are North America or Europe. They're pretty well off, on average. About 1.3bn are in China. They are dramatically richer than they were 20 years ago. And that's true of most of Asia.
Median incomes and lifestyles have increased extraordinarily in the last 20 years. I mean some places, like rural Russia, have gone backwards. But most places are dramatically richer than they were.
Okay. Uganda is well below average, so yes. Brazil isn't far below it, though.
I've been hunting for the world's median person. Listing countries by PPPpc and then going down the list until half the world's population is accounted for, a disgustingly quick and dirty method that has all sorts of things wrong with it, gives the following 7 countries within 10 of the median and with population >10 million:
In Asia...........Indonesia, Sri Lanka In America.....Brazil, Colombia, Peru In Africa........Algeria, South Africa
China is above, India below, by this Q&D metric.
GDP per capita will tend to exaggerate the personal incomes of some resource countries. Take Peru: in large part thanks to a massive LNG discovery, its GDP has quadrupled in the last twenty years from $60bn to $240bn.
Yet I'll bet you that relatively little of that has fed through to incomes.
By contrast, look at Bangladesh. Two decades ago, incomes were less than $2,000 per year, and incomes were half those of Pakistan. Famines were a regular occurance, as were natural disasters that would hammer them.
And now it's a lot richer than Pakistan on a per capita basis, and may well have overtaken India.
So my take on climate change is this: everywhere in the world is going to get hotter and hotter… apart from the UK which will, uniquely, contrive to get greyer and rainier and even more dismal
Oh well. At least it should discourage the boat people
I agree. Our winters are getting warmer but wetter and more miserable and our summers warmer but cloudier. Still one of the most dismal climates for its latitude in the world is Lima which despite being in the Tropics manages to be under constant grey 8 months of the year.
Indeed. Lima manages to be climatically way more miserable than anywhere in the UK, which is quite a feat given its location. It always annoys me when I read history books or guide books that reference Lima and don’t mention this fairly notable affliction
Why the fuck did the Spanish build their capital there? The incans very sensibly chose sunny and refreshing Cusco
Possibly the worst “place” I have ever been - in terms of climate meeting geography - is the desert north of Lima. The Sechura. It’s a dismal grey sand desert, strewn with trash, and cursed with that same cruel and depressing climate - chilly grey cloud like Glasgow but without the chirpy locals
Also shit food and a history of urgent child sacrifice
One of the worst drives of my life was in the desert north of lima stuck with a mad driver driving like a maniac on a narrow road and swerving at the last minute to avoid oncoming traffic. And yes its bleak and the locals are miserable.
I took this photo on a beach in deserty northern Peru as it seemed to summarise the whole place
However the little colonial towns do have a certain charm, under those sparkling blue skies
Very Quantum of Solace.
Imagine living on that street in the second photo. Wake up, put your head out of the door, hang yourself
Most people in the world live somewhere like that. It's Median Street, Planet Earth, 2024. Most people have never stayed in a hotel, owned a car, held a year's income in a bank account, etc. Yet suicide is more frequent in the USA and Canada than it is in Latin America. And the curious thing: that's with "Latin America" defined as Hispanophone and Lusophone America. There is a single country in South America that tops USA and Canada for self-topping: Guyana.
As someone who constantly travels the world, I can (happily) reassure you that most people do NOT live on a street similar to that street in the 2nd photo: ie on a dirt road, in a concrete shack, in a shitty town in the middle of an awful foggy desert. Northern Peru, as we have established, is unusually hideous
The global median person probably lives in a concrete apartment in a large city in India or China. Not beautiful, but not terrible
Hotels and resorts aren't the world. Two thirds of the population of India, the world's most populous country, live outside of cities and so presumably not in apartments. That photo could easily be from many places in Brazil (even 30 miles from the capital) or Uganda (ditto). China, yes, mostly apartments - only a third live outside of cities. Depends how you define median. And cities. I would say they're in a shitty small town somewhere in the third world. Smartphones being the hard drug that's available.
Well, if you think Uganda (GDP per capita $883) is anything like average for the world, you're insane.
There are 8 billion people on Planet Earth (give or take).
About 1.4bn of them are North America or Europe. They're pretty well off, on average. About 1.3bn are in China. They are dramatically richer than they were 20 years ago. And that's true of most of Asia.
Median incomes and lifestyles have increased extraordinarily in the last 20 years. I mean some places, like rural Russia, have gone backwards. But most places are dramatically richer than they were.
Okay. Uganda is well below average, so yes. Brazil isn't far below it, though.
I've been hunting for the world's median person. Listing countries by PPPpc and then going down the list until half the world's population is accounted for, a disgustingly quick and dirty method that has all sorts of things wrong with it, gives the following 7 countries within 10 of the median and with population >10 million:
In Asia...........Indonesia, Sri Lanka In America.....Brazil, Colombia, Peru In Africa........Algeria, South Africa
China is above, India below, by this Q&D metric.
The 2 of the 7 with closest to the median murder rate are:
Algeria, Sri Lanka
Our person may be in a small town in one of those. Where would you locate him or her?
Well I I’ve been to all of those countries - except Algeria. Can anyone else on here say that?
Indeed I am right now in a fairly average coastal city in Colombia
And I can report that none of these countries has single storey concrete shacks by dirt roads as a median living experience for their citizens
So my take on climate change is this: everywhere in the world is going to get hotter and hotter… apart from the UK which will, uniquely, contrive to get greyer and rainier and even more dismal
Oh well. At least it should discourage the boat people
I agree. Our winters are getting warmer but wetter and more miserable and our summers warmer but cloudier. Still one of the most dismal climates for its latitude in the world is Lima which despite being in the Tropics manages to be under constant grey 8 months of the year.
Indeed. Lima manages to be climatically way more miserable than anywhere in the UK, which is quite a feat given its location. It always annoys me when I read history books or guide books that reference Lima and don’t mention this fairly notable affliction
Why the fuck did the Spanish build their capital there? The incans very sensibly chose sunny and refreshing Cusco
Possibly the worst “place” I have ever been - in terms of climate meeting geography - is the desert north of Lima. The Sechura. It’s a dismal grey sand desert, strewn with trash, and cursed with that same cruel and depressing climate - chilly grey cloud like Glasgow but without the chirpy locals
Also shit food and a history of urgent child sacrifice
One of the worst drives of my life was in the desert north of lima stuck with a mad driver driving like a maniac on a narrow road and swerving at the last minute to avoid oncoming traffic. And yes its bleak and the locals are miserable.
I took this photo on a beach in deserty northern Peru as it seemed to summarise the whole place
However the little colonial towns do have a certain charm, under those sparkling blue skies
Very Quantum of Solace.
Imagine living on that street in the second photo. Wake up, put your head out of the door, hang yourself
Most people in the world live somewhere like that. It's Median Street, Planet Earth, 2024. Most people have never stayed in a hotel, owned a car, held a year's income in a bank account, etc. Yet suicide is more frequent in the USA and Canada than it is in Latin America. And the curious thing: that's with "Latin America" defined as Hispanophone and Lusophone America. There is a single country in South America that tops USA and Canada for self-topping: Guyana.
As someone who constantly travels the world, I can (happily) reassure you that most people do NOT live on a street similar to that street in the 2nd photo: ie on a dirt road, in a concrete shack, in a shitty town in the middle of an awful foggy desert. Northern Peru, as we have established, is unusually hideous
The global median person probably lives in a concrete apartment in a large city in India or China. Not beautiful, but not terrible
Hotels and resorts aren't the world. Two thirds of the population of India, the world's most populous country, live outside of cities and so presumably not in apartments. That photo could easily be from many places in Brazil (even 30 miles from the capital) or Uganda (ditto). China, yes, mostly apartments - only a third live outside of cities. Depends how you define median. And cities. I would say they're in a shitty small town somewhere in the third world. Smartphones being the hard drug that's available.
Well, if you think Uganda (GDP per capita $883) is anything like average for the world, you're insane.
There are 8 billion people on Planet Earth (give or take).
About 1.4bn of them are North America or Europe. They're pretty well off, on average. About 1.3bn are in China. They are dramatically richer than they were 20 years ago. And that's true of most of Asia.
Median incomes and lifestyles have increased extraordinarily in the last 20 years. I mean some places, like rural Russia, have gone backwards. But most places are dramatically richer than they were.
Okay. Uganda is well below average, so yes. Brazil isn't far below it, though.
I've been hunting for the world's median person. Listing countries by PPPpc and then going down the list until half the world's population is accounted for, a disgustingly quick and dirty method that has all sorts of things wrong with it, gives the following 7 countries within 10 of the median and with population >10 million:
In Asia...........Indonesia, Sri Lanka In America.....Brazil, Colombia, Peru In Africa........Algeria, South Africa
China is above, India below, by this Q&D metric.
GDP per capita will tend to exaggerate the personal incomes of some resource countries. Take Peru: in large part thanks to a massive LNG discovery, its GDP has quadrupled in the last twenty years from $60bn to $240bn.
Yet I'll bet you that relatively little of that has fed through to incomes.
By contrast, look at Bangladesh. Two decades ago, incomes were less than $2,000 per year, and incomes were half those of Pakistan. Famines were a regular occurance, as were natural disasters that would hammer them.
And now it's a lot richer than Pakistan on a per capita basis, and may well have overtaken India.
My guess is 98.3% of humanity couldn’t solve that imperial currency puzzle, at least not without some research into LSD, so in that respect AI is very very human
So my take on climate change is this: everywhere in the world is going to get hotter and hotter… apart from the UK which will, uniquely, contrive to get greyer and rainier and even more dismal
Oh well. At least it should discourage the boat people
I agree. Our winters are getting warmer but wetter and more miserable and our summers warmer but cloudier. Still one of the most dismal climates for its latitude in the world is Lima which despite being in the Tropics manages to be under constant grey 8 months of the year.
Indeed. Lima manages to be climatically way more miserable than anywhere in the UK, which is quite a feat given its location. It always annoys me when I read history books or guide books that reference Lima and don’t mention this fairly notable affliction
Why the fuck did the Spanish build their capital there? The incans very sensibly chose sunny and refreshing Cusco
Possibly the worst “place” I have ever been - in terms of climate meeting geography - is the desert north of Lima. The Sechura. It’s a dismal grey sand desert, strewn with trash, and cursed with that same cruel and depressing climate - chilly grey cloud like Glasgow but without the chirpy locals
Also shit food and a history of urgent child sacrifice
One of the worst drives of my life was in the desert north of lima stuck with a mad driver driving like a maniac on a narrow road and swerving at the last minute to avoid oncoming traffic. And yes its bleak and the locals are miserable.
I took this photo on a beach in deserty northern Peru as it seemed to summarise the whole place
However the little colonial towns do have a certain charm, under those sparkling blue skies
Very Quantum of Solace.
Imagine living on that street in the second photo. Wake up, put your head out of the door, hang yourself
Most people in the world live somewhere like that. It's Median Street, Planet Earth, 2024. Most people have never stayed in a hotel, owned a car, held a year's income in a bank account, etc. Yet suicide is more frequent in the USA and Canada than it is in Latin America. And the curious thing: that's with "Latin America" defined as Hispanophone and Lusophone America. There is a single country in South America that tops USA and Canada for self-topping: Guyana.
As someone who constantly travels the world, I can (happily) reassure you that most people do NOT live on a street similar to that street in the 2nd photo: ie on a dirt road, in a concrete shack, in a shitty town in the middle of an awful foggy desert. Northern Peru, as we have established, is unusually hideous
The global median person probably lives in a concrete apartment in a large city in India or China. Not beautiful, but not terrible
Hotels and resorts aren't the world. Two thirds of the population of India, the world's most populous country, live outside of cities and so presumably not in apartments. That photo could easily be from many places in Brazil (even 30 miles from the capital) or Uganda (ditto). China, yes, mostly apartments - only a third live outside of cities. Depends how you define median. And cities. I would say they're in a shitty small town somewhere in the third world. Smartphones being the hard drug that's available.
Well, if you think Uganda (GDP per capita $883) is anything like average for the world, you're insane.
There are 8 billion people on Planet Earth (give or take).
About 1.4bn of them are North America or Europe. They're pretty well off, on average. About 1.3bn are in China. They are dramatically richer than they were 20 years ago. And that's true of most of Asia.
Median incomes and lifestyles have increased extraordinarily in the last 20 years. I mean some places, like rural Russia, have gone backwards. But most places are dramatically richer than they were.
Okay. Uganda is well below average, so yes. Brazil isn't far below it, though.
I've been hunting for the world's median person. Listing countries by PPPpc and then going down the list until half the world's population is accounted for, a disgustingly quick and dirty method that has all sorts of things wrong with it, gives the following 7 countries within 10 of the median and with population >10 million:
In Asia...........Indonesia, Sri Lanka In America.....Brazil, Colombia, Peru In Africa........Algeria, South Africa
China is above, India below, by this Q&D metric.
GDP per capita will tend to exaggerate the personal incomes of some resource countries. Take Peru: in large part thanks to a massive LNG discovery, its GDP has quadrupled in the last twenty years from $60bn to $240bn.
Yet I'll bet you that relatively little of that has fed through to incomes.
By contrast, look at Bangladesh. Two decades ago, incomes were less than $2,000 per year, and incomes were half those of Pakistan. Famines were a regular occurance, as were natural disasters that would hammer them.
And now it's a lot richer than Pakistan on a per capita basis, and may well have overtaken India.
I think Rishi has got it wrong too to let it go on further .
I’m not convinced by this, “so called” confirmation. It wasn’t a planned announcement. He didn’t plan to say it or else he wouldn’t have dodged the question twenty times first. It seems an odd way to break the decision, pressed more than several times in a regional interview, and then slipping it out in frustration…
I’m also highly suspicious how the media then leapt on it, without thinking the same as what I’m thinking. He will be asked to clarify and confirm by whoever’s turn it is in the Westminster press pool tomorrow, so let’s see how unequivocal Sunak, and the briefing from The Mouth of Sauron is.
Let’s see what Friday brings 😉
Have I missed something? What has Sunak said?
In a milking shed in the south West Country, Sunak said he won’t call an election until horsecorrectbattery4 is unbanned. Though at the end of the interview conceded, if there’s a spam trap, there’s likely a very naughty word trap too, and if he was a betting man he would have a good punt of the poster most likely to spring it.
Nice to have friends in high places? Make the most of it, he’s probably only got a matter of weeks left in the job.
I’ve just been to my local supermarket. It’s a testament to globalisation
Now to be fair I am in an affluent seaside neighborhood. Nonetheless this is a pretty average Colombian city. It’s not famous like Cartagena (so no tourist income), it’s not important like bogota. It’s not big like Medellin
It’s a city of 500,000 people quite near the troubled Venezuelan border in a country with a per capita income of $7000
At first glance the supermarket could easily be in Spain or Croatia. Or a richer bit of wales. The cheese selection is better than most supermarkets in America. The wine is from all over. - but lots of chile and Argentinian. The veg and fruit selection is great and fresh. The bread is good. The hard liquor is fine - bought Bombay sapphire but could have bought Tanqueray or Hendricks
Wide selection of European olive oils. Great chocolates. Obviously impressive array of coffees ground, beans, raw, all forms. About ten different balsamic vinegars. Dijon mustard. Loads of beers
This - I repeat - is a nice supermarket in a nice but not outrageously wealthy neighborhood in an average and quite remote Colombian city
Globalisation and capitalism are phenomenal things
Meta's chief AI wallah tells Musk he is wrong. Sorry @Leon
Yann LeCun
@ylecun · 4h No. If it were the case, we would have AI systems that could teach themselves to drive a car in 20 hours of practice, like any 17 year-old.
But we still don't have fully autonomous, reliable self-driving, even though we (you) have millions of hours of *labeled* training data.
A 17 year old whose prefrontal cortex isn't even fully developed can learn to drive with 20 hours of training.
AI can't learn to drive with millions of hours of training.
The idea AI is smarter than humans is facetious bullshit. It might know more, but its not smarter.
Current-generation AIs don't even know what year it is. Facts are not exactly their strong point. But they haven't been designed to do so.
AI won't be smart until it can do stuff it hasn't been designed to do, simply because it wants to do it, without anyone prompting it to do so.
Like any toddler can.
Er, GPT3 did exactly that. No one taught it to draw. No one realised it had that capability. Then someone asked it to draw - and it did
“AI is just an evolution of stuff we already have, it’s not some great new thing.”
“The internet is exactly like a vegetable spiralizer”
“I can tell you for a fact, AI will provide some efficiency gains and that will be the end of it.”
Ummm:
"Er, GPT3 did exactly that. No one taught it to draw. No one realised it had that capability. Then someone asked it to draw - and it did "
I asked ChatGPT (version 4) to assess the veracity of that claim and it said:
The statement you're referring to seems to mix up a few details about AI development and capabilities. GPT-3, developed by OpenAI, is a language model known for its ability to understand and generate human-like text based on the input it receives. It was not designed with the capability to create visual art or drawings; its strengths lie in text generation, comprehension, translation, and other language-based tasks.
The confusion might come from the development of other AI models that are capable of generating images, such as DALL·E (also developed by OpenAI), which was specifically designed to create images from textual descriptions. Unlike GPT-3, DALL·E was intentionally developed and trained to translate text descriptions into visual content. So, when DALL·E is asked to draw something based on a text prompt, it can generate images that match the prompt to a remarkable degree of creativity and accuracy.
To summarize, the statement is not accurate regarding GPT-3, as GPT-3's capabilities are centered around text. The ability to generate drawings or images from text descriptions relates to other AI models like DALL·E, which were specifically designed for that purpose.
I think Rishi has got it wrong too to let it go on further .
I’m not convinced by this, “so called” confirmation. It wasn’t a planned announcement. He didn’t plan to say it or else he wouldn’t have dodged the question twenty times first. It seems an odd way to break the decision, pressed more than several times in a regional interview, and then slipping it out in frustration…
I’m also highly suspicious how the media then leapt on it, without thinking the same as what I’m thinking. He will be asked to clarify and confirm by whoever’s turn it is in the Westminster press pool tomorrow, so let’s see how unequivocal Sunak, and the briefing from The Mouth of Sauron is.
I’ve just been to my local supermarket. It’s a testament to globalisation
Now to be fair I am in an affluent seaside neighborhood. Nonetheless this is a pretty average Colombian city. It’s not famous like Cartagena (so no tourist income), it’s not important like bogota. It’s not big like Medellin
It’s a city of 500,000 people quite near the troubled Venezuelan border in a country with a per capita income of $7000
At first glance the supermarket could easily be in Spain or Croatia. Or a richer bit of wales. The cheese selection is better than most supermarkets in America. The wine is from all over. - but lots of chile and Argentinian. The veg and fruit selection is great and fresh. The bread is good. The hard liquor is fine - bought Bombay sapphire but could have bought Tanqueray or Hendricks
Wide selection of European olive oils. Great chocolates. Obviously impressive array of coffees ground, beans, raw, all forms. About ten different balsamic vinegars. Dijon mustard. Loads of beers
This - I repeat - is a nice supermarket in a nice but not outrageously wealthy neighborhood in an average and quite remote Colombian city
Globalisation and capitalism are phenomenal things
Have you spoken with any Colombians about the Colombian caste system yet? There are 6 levels, with number 6 at the top. That supermarket sounds like 4, possibly 5.
Meta's chief AI wallah tells Musk he is wrong. Sorry @Leon
Yann LeCun
@ylecun · 4h No. If it were the case, we would have AI systems that could teach themselves to drive a car in 20 hours of practice, like any 17 year-old.
But we still don't have fully autonomous, reliable self-driving, even though we (you) have millions of hours of *labeled* training data.
A 17 year old whose prefrontal cortex isn't even fully developed can learn to drive with 20 hours of training.
AI can't learn to drive with millions of hours of training.
The idea AI is smarter than humans is facetious bullshit. It might know more, but its not smarter.
Current-generation AIs don't even know what year it is. Facts are not exactly their strong point. But they haven't been designed to do so.
AI won't be smart until it can do stuff it hasn't been designed to do, simply because it wants to do it, without anyone prompting it to do so.
Like any toddler can.
Er, GPT3 did exactly that. No one taught it to draw. No one realised it had that capability. Then someone asked it to draw - and it did
“AI is just an evolution of stuff we already have, it’s not some great new thing.”
“The internet is exactly like a vegetable spiralizer”
“I can tell you for a fact, AI will provide some efficiency gains and that will be the end of it.”
Ummm:
"Er, GPT3 did exactly that. No one taught it to draw. No one realised it had that capability. Then someone asked it to draw - and it did "
I asked ChatGPT (version 4) to assess the veracity of that claim and it said:
The statement you're referring to seems to mix up a few details about AI development and capabilities. GPT-3, developed by OpenAI, is a language model known for its ability to understand and generate human-like text based on the input it receives. It was not designed with the capability to create visual art or drawings; its strengths lie in text generation, comprehension, translation, and other language-based tasks.
The confusion might come from the development of other AI models that are capable of generating images, such as DALL·E (also developed by OpenAI), which was specifically designed to create images from textual descriptions. Unlike GPT-3, DALL·E was intentionally developed and trained to translate text descriptions into visual content. So, when DALL·E is asked to draw something based on a text prompt, it can generate images that match the prompt to a remarkable degree of creativity and accuracy.
To summarize, the statement is not accurate regarding GPT-3, as GPT-3's capabilities are centered around text. The ability to generate drawings or images from text descriptions relates to other AI models like DALL·E, which were specifically designed for that purpose.
That’s interestingly wrong
It’s 8pm in Colombia and I need to eat, but image creation was an emergent property of the early GPTs. The first Dall-e was a remake of GPT3 I think
I’ve just been to my local supermarket. It’s a testament to globalisation
Now to be fair I am in an affluent seaside neighborhood. Nonetheless this is a pretty average Colombian city. It’s not famous like Cartagena (so no tourist income), it’s not important like bogota. It’s not big like Medellin
It’s a city of 500,000 people quite near the troubled Venezuelan border in a country with a per capita income of $7000
At first glance the supermarket could easily be in Spain or Croatia. Or a richer bit of wales. The cheese selection is better than most supermarkets in America. The wine is from all over. - but lots of chile and Argentinian. The veg and fruit selection is great and fresh. The bread is good. The hard liquor is fine - bought Bombay sapphire but could have bought Tanqueray or Hendricks
Wide selection of European olive oils. Great chocolates. Obviously impressive array of coffees ground, beans, raw, all forms. About ten different balsamic vinegars. Dijon mustard. Loads of beers
This - I repeat - is a nice supermarket in a nice but not outrageously wealthy neighborhood in an average and quite remote Colombian city
Globalisation and capitalism are phenomenal things
Have you spoken with any Colombians about the Colombian caste system yet? There are 6 levels, with number 6 at the top. That supermarket sounds like 4, possibly 5.
I’ve said I’m in a richer burb. I don’t think the average Colombian is buying Camembert or Tanqueray Ten
Nonetheless the locals are not creeping into this supermarket in awe at the wealth either
I’ve been all over Colombia these last two weeks - capital, jungle, Medellin, Cartagena now Santa Marta. I’ve been on buses and in taxis and in minivans
I’ve not seen a single place that looks as poor and shit as that road in Peru which triggered this whole argument. So I am right and you are wrong
Meta's chief AI wallah tells Musk he is wrong. Sorry @Leon
Yann LeCun
@ylecun · 4h No. If it were the case, we would have AI systems that could teach themselves to drive a car in 20 hours of practice, like any 17 year-old.
But we still don't have fully autonomous, reliable self-driving, even though we (you) have millions of hours of *labeled* training data.
A 17 year old whose prefrontal cortex isn't even fully developed can learn to drive with 20 hours of training.
AI can't learn to drive with millions of hours of training.
The idea AI is smarter than humans is facetious bullshit. It might know more, but its not smarter.
Current-generation AIs don't even know what year it is. Facts are not exactly their strong point. But they haven't been designed to do so.
AI won't be smart until it can do stuff it hasn't been designed to do, simply because it wants to do it, without anyone prompting it to do so.
Like any toddler can.
Er, GPT3 did exactly that. No one taught it to draw. No one realised it had that capability. Then someone asked it to draw - and it did
“AI is just an evolution of stuff we already have, it’s not some great new thing.”
“The internet is exactly like a vegetable spiralizer”
“I can tell you for a fact, AI will provide some efficiency gains and that will be the end of it.”
Ummm:
"Er, GPT3 did exactly that. No one taught it to draw. No one realised it had that capability. Then someone asked it to draw - and it did "
I asked ChatGPT (version 4) to assess the veracity of that claim and it said:
The statement you're referring to seems to mix up a few details about AI development and capabilities. GPT-3, developed by OpenAI, is a language model known for its ability to understand and generate human-like text based on the input it receives. It was not designed with the capability to create visual art or drawings; its strengths lie in text generation, comprehension, translation, and other language-based tasks.
The confusion might come from the development of other AI models that are capable of generating images, such as DALL·E (also developed by OpenAI), which was specifically designed to create images from textual descriptions. Unlike GPT-3, DALL·E was intentionally developed and trained to translate text descriptions into visual content. So, when DALL·E is asked to draw something based on a text prompt, it can generate images that match the prompt to a remarkable degree of creativity and accuracy.
To summarize, the statement is not accurate regarding GPT-3, as GPT-3's capabilities are centered around text. The ability to generate drawings or images from text descriptions relates to other AI models like DALL·E, which were specifically designed for that purpose.
That’s interestingly wrong
It’s 8pm in Colombia and I need to eat, but image creation was an emergent property of the early GPTs. The first Dall-e was a remake of GPT3 I think
That's simply not true. I've explained how Dall-E has worked many times before. It's brilliant in its simplicity, and like ChatGPT it uses a neural net.
But Dall-E did not come out of GPT. They are two, entirely different systems.
(With the proviso that the latest iterations connect them: ChatGPT 4 is capable of feeding Dall-E, and other systems. But that does not mean that Dall-E was in any way an emergent product of GPT.)
Comments
That should be more LOL then!!!
Can the cones hotline be bested?
The Japanese are polite. Obsessively so
Parisians are rude. Yep
Like any toddler can.
Calculators can do calculations instantly and accurately that people would struggle to do mentally. When calculators arose people were saying how bad that was for all sorts of reasons etc
But calculators are a tool we can just use everyday now and take it for granted. Just like the internet is, just like electricity is, and just like AI will be going forwards too.
Smart is knowing how to use those tools, not being able to outprocess the tools in what they've been designed to do.
My calculator can do calculations I can't in my head, but I'm smarter than the calculator as I know how to use it, not the other way around. Same with AI.
He’s one of the godfathers of AI but he’s well past it. He’s made loads of embarrassing mistakes in recent years
Here’s one. He recently claimed that LLMs could not understand the basic physics of an object placed on a table
Then GPT3.5 did exactly that
https://youtu.be/sWF6SKfjtoU?si=IVbu3nOFU4tg6ZqQ
I think he just doesn’t grasp how fast and why AI is developing but as a proud man with a remarkable CV he’s reluctant to retire and accept his redundancy
“AI is just an evolution of stuff we already have, it’s not some great new thing.”
“The internet is exactly like a vegetable spiralizer”
“I can tell you for a fact, AI will provide some efficiency gains and that will be the end of it.”
“AI” is really really really really really really really really not “just like the invention of the pocket calculator”
https://www.bbc.co.uk/news/uk-scotland-edinburgh-east-fife-68308809
Guess what block chain was overhyped bollocks so are llms but it gives no nothing idiot journalists something to froth about and they will be just as wrong now as they were about the importance of block chain or what 3 words. The reason we despise journalists is they write shit about stuff while having no understanding and getting it wrong. I trust politicians more than journalists and I have no faith whatsoever in politicians
Messrs Gove, Johnson ...
"... someone asked it to draw ..."
So no, that's not unprompted. Someone gave it the prompt and it did it.
AI 100% is just an evolution of what we already have.
I never said the internet is exactly like a vegetable spiraliser, you did, I said its a tool and it is. Also tools are the internet, calculators and vegetable spiralisers, yes. Corkscrews too.
Different tools have different uses.
Its a tool to do quickly what the user wants it to do.
Its not a sentient, self-actualising body that is doing what it wants to do all by itself.
https://writings.stephenwolfram.com/2023/02/what-is-chatgpt-doing-and-why-does-it-work/
Atmo you seem like a priest without theology.
This was in the late 80s
Will this torture of the Red Wall never stop?
He's the sort where if you go to tenerife, he has to go to elevenerife.
My father and I are both lifelong Republicans. He got aboard the “Trump Train” and I waved farewell from the station.
The sad thing is, I’ve loved talking politics with him all my life. Insightful, well-reasoned, a wealth of insight even if we didn’t agree on everything.
Now, it’s all slogans and empty bullshit bromides. Jokes about pronouns and other intellect-light culture war nonsense.
MAGA didn’t just make him angrier. It made him stupider, too. Perhaps its greatest accomplishment.
10:26 PM · Mar 14, 2024
·
794
Views
https://x.com/Ibram_Gaunt77/status/1768403239160983637?s=20
... However!
... Then a blitz of....
My only few IT bets down the years have been Linux, The Internet, 'Virtualisation' and 'Mobile'. I've been on a bit of a streak - but 'AI' is giving similar twitchy nose vibes,
NOW he gets it!
The lager-drinking device used on the catastrophist is rather obvious and grossly overdone, though.
could be in REAL trouble!'
@Nigel_Farage
: 'Hope Not Hate's 'State of Hate Report' mentions me 56 times (more than any other individual), GB News is mentioned 'page after page', and people could use the word hate 'as a cover to shut down anyone they disagree with'.
https://x.com/GBNEWS/status/1768376806691020949?s=20
This is the RefUK candidate in my Assembly constituency: https://www.reformparty.uk/gla-barnet-and-camden He used to be in the Conservatives and, before that, supported the BNP! However, the BNP in question was the Bangladesh Nationalist Party.
I hesitate to insult fellow PBers by revealing the answer to be two and ninepence - you'll have worked that out for ourselves, no doubt. But I'd be interested to know if AI could do it as quickly.
LAB: 44% (=)
CON: 22% (-1)
RFM: 12% (+1)
LDM: 11% (+1)
GRN: 5% (-1)
SNP: 3% (=)
Via
@techneUK
, 13-14 Mar.
Changes w/ 6-7 Mar.
Cricklade & Latton (Wiltshire) Council By-Election Result:
🔶 LDM: 76.0% (+13.9)
🌳 CON: 18.7% (-19.2)
🌹 LAB: 3.3% (New)
🌍 GEN: 2.0% (New)
Liberal Democrat HOLD.
Changes w/ 2021.
No RefUK as usual but LD ahead of national polling again (local by-election so don't read too much into that)
Nah. He doesn’t deserve fair
Castle (Lancaster) Council By-Election Result:
🌍 GRN: 65.0% (+6.4)
🌹 LAB: 26.3 (-3.0)
🌳 CON: 5.3% (-1.8)
🔶 LDM: 3.3% (-1.7)
Green Party HOLD.
Changes w/ 2023.
It is very difficult to judge how RefUK will do when they never stand. Frustrating. But following on from a previous comment, a good sign for the Greens. I find it hard to believe that come the election, RefUK will significantly outpoll them when there is no evidence of it on the ground.
However I doubt it swings any seats, apart from maybe some of the Barnet seats back to the Tories if Labour push too hard for a ceasefire, or Bradford and Tower Hamlets in addition to Rochdale to Galloway if Labour are seen as too pro Israel
Both Johnson and Macron, for all their myriad other failings (Johnson more than Macron) were absolutely right with the way they handled the lead up to and start of the invasion.
Adding in shillings where appropriate helped though Gemini for some reason decided to round up to the nearest shilling and GPT3.5 almost got it and then forgot how many pennies were in a shilling to come up with a very confident and wrong 2 shillings and 4 pence). GPT4 did get it completely right with an python code that I hadn't asked for and an odd moral lesson added.
Jack has 3 shillings and 7 pence, while Jill has 4 shillings and 8 pence. Let's convert everything into pence first for ease of calculation, then add them up, and finally divide the total by three (since they have three children) to figure out how much each child can receive. After that, we'll convert the amount back into shillings and pence for each child.
Let's calculate:
Jack and Jill can give each child 2 shillings and 9 pence. It's a tidy sum, considering the era of shillings and pence! Distributing their savings equally among their children shows a fair and practical approach to pocket money, ensuring each child receives a portion to spend or save as they see fit.
# Conversion rates
pence_per_shilling = 12
# Jack's money in pence
jack_shillings = 3
jack_pence = 7
jack_total_pence = jack_shillings * pence_per_shilling + jack_pence
# Jill's money in pence
jill_shillings = 4
jill_pence = 8
jill_total_pence = jill_shillings * pence_per_shilling + jill_pence
# Total money in pence
total_pence = jack_total_pence + jill_total_pence
# Divide by three children
per_child_pence = total_pence / 3
# Convert back to shillings and pence for each child
child_shillings, child_pence = divmod(per_child_pence, pence_per_shilling)
(child_shillings, child_pence)
If those were pounds you're talking about, then that's another 33 pence each with a penny left over.
So I'm going £2.48 each with a penny leftover.
So £2.38 that they got is the right answer (I did a typo writing £2.48).
£2.05 each plus 33p each from the extra pound.
That deserves sympathy!
Oddly for the question on "In 1922 how did the Seine cause 86 people to drown?" GPT3 came up with what seems like a completely made up story of a river paddle steamer full of Parisian daytrippers colliding with a tugboat with massive loss of life. Just now I did the same question to GPT4 to get the same sort of answer, just with more detail, but still completely wrong.
Pence have gone with pounds for my entire lifetime, let alone GPT's lifetime.
I've been hunting for the world's median person. Listing countries by PPPpc and then going down the list until half the world's population is accounted for, a disgustingly quick and dirty method that has all sorts of things wrong with it, gives the following 7 countries within 10 of the median and with population >10 million:
In Asia...........Indonesia, Sri Lanka
In America.....Brazil, Colombia, Peru
In Africa........Algeria, South Africa
China is above, India below, by this Q&D metric.
Algeria, Sri Lanka
Our person may be in a small town in one of those.
Where would you locate him or her?
I’m also highly suspicious how the media then leapt on it, without thinking the same as what I’m thinking. He will be asked to clarify and confirm by whoever’s turn it is in the Westminster press pool tomorrow, so let’s see how unequivocal Sunak, and the no 10 briefing is.
We will at least know tomorrow either way, if they don’t row back on media interesting it’s not May second, then probably it’s not May 2nd. However if they do row back on it to the will he won’t he position again, that will mean it’s definitely May 2nd.
Let’s see what Friday brings 😉
Yet I'll bet you that relatively little of that has fed through to incomes.
By contrast, look at Bangladesh. Two decades ago, incomes were less than $2,000 per year, and incomes were half those of Pakistan. Famines were a regular occurance, as were natural disasters that would hammer them.
And now it's a lot richer than Pakistan on a per capita basis, and may well have overtaken India.
This podcast - https://www.npr.org/sections/money/2013/05/14/184019151/episode-458-bangladeshs-t-shirt-economy - is ten years old, but gives you a really good feel for how Bangladesh has managed to do it. It also gives you a really good feel for how people in these countries are thriving.
Edit to add: everyone, particularly opponents of globalisation, should listen to the podcast. It's absolutely fascinating.
Indeed I am right now in a fairly average coastal city in Colombia
And I can report that none of these countries has single storey concrete shacks by dirt roads as a median living experience for their citizens
So I’m right and you’re wrong
Bangladesh is a remarkable and unsung success story
Nice to have friends in high places? Make the most of it, he’s probably only got a matter of weeks left in the job.
Now to be fair I am in an affluent seaside neighborhood. Nonetheless this is a pretty average Colombian city. It’s not famous like Cartagena (so no tourist income), it’s not important like bogota. It’s not big like Medellin
It’s a city of 500,000 people quite near the troubled Venezuelan border in a country with a per capita income of $7000
At first glance the supermarket could easily be in Spain or Croatia. Or a richer bit of wales. The cheese selection is better than most supermarkets in America. The wine is from all over. - but lots of chile and Argentinian. The veg and fruit selection is great and fresh. The bread is good. The hard liquor is fine - bought Bombay sapphire but could have bought Tanqueray or Hendricks
Wide selection of European olive oils. Great chocolates. Obviously impressive array of coffees ground, beans, raw, all forms. About ten different balsamic vinegars. Dijon mustard. Loads of beers
This - I repeat - is a nice supermarket in a nice but not outrageously wealthy neighborhood in an average and quite remote Colombian city
Globalisation and capitalism are phenomenal things
"Er, GPT3 did exactly that. No one taught it to draw. No one realised it had that capability. Then someone asked it to draw - and it did "
I asked ChatGPT (version 4) to assess the veracity of that claim and it said:
The statement you're referring to seems to mix up a few details about AI development and capabilities. GPT-3, developed by OpenAI, is a language model known for its ability to understand and generate human-like text based on the input it receives. It was not designed with the capability to create visual art or drawings; its strengths lie in text generation, comprehension, translation, and other language-based tasks.
The confusion might come from the development of other AI models that are capable of generating images, such as DALL·E (also developed by OpenAI), which was specifically designed to create images from textual descriptions. Unlike GPT-3, DALL·E was intentionally developed and trained to translate text descriptions into visual content. So, when DALL·E is asked to draw something based on a text prompt, it can generate images that match the prompt to a remarkable degree of creativity and accuracy.
To summarize, the statement is not accurate regarding GPT-3, as GPT-3's capabilities are centered around text. The ability to generate drawings or images from text descriptions relates to other AI models like DALL·E, which were specifically designed for that purpose.
There are 6 levels, with number 6 at the top. That supermarket sounds like 4, possibly 5.
https://en.wikipedia.org/wiki/Social_class_in_Colombia
It’s 8pm in Colombia and I need to eat, but image creation was an emergent property of the early GPTs. The first Dall-e was a remake of GPT3 I think
Nonetheless the locals are not creeping into this supermarket in awe at the wealth either
I’ve been all over Colombia these last two weeks - capital, jungle, Medellin, Cartagena now Santa Marta. I’ve been on buses and in taxis and in minivans
I’ve not seen a single place that looks as poor and shit as that road in Peru which triggered this whole argument. So I am right and you are wrong
But Dall-E did not come out of GPT. They are two, entirely different systems.
(With the proviso that the latest iterations connect them: ChatGPT 4 is capable of feeding Dall-E, and other systems. But that does not mean that Dall-E was in any way an emergent product of GPT.)