Neural networks were developed to mimic visual cortex neurons. We're still at a crude place of simulation, but I wouldn't say AI has _nothing_ to tell us about language. Attention models for example raise interesting questions about how we process language.
Well... "X has something to say about Y" can have a pretty generous interpretation. What's your example? usually these kinds of claims come from people who have not delved into topics of cognition and language and how complicated they can be. LLMs may say something about abstract notion of "language" in general but Chomsky's completely correct they don't teach us anything about how humans learn language... maybe if they overcome poverty of stimulus one day, that could be a first step.
@@iranjackheelson My point was that attending to words/subwords as a pathway to gather wider meaning, along with the emergent generalizable quality of networks of neurons, has a lot to teach us about the roots of intelligence. Even if as a foil or opposite, I think throwing out that modality of learning would a mistake. Certainly I think saying deep learning has *nothing* to teach us about complex learning functions like language is a very bold claim that you need to defend with data and logic. It's not a suitable throw away line.
@@dan_taninecz_geopol I got your point but again just goes back my point that you don't seem to understand how language works in human context. what you've pointed out are aspects of neuralnet that have something to teach us about anything that's associative/statistical learning based, not just language. sure, there's some sense language learning and usage is associative, but that explains negligibly miniscule aspect of how humans use language. the burden of proof is on you to say there's more explanatory power beyond the simple associative aspect
@@iranjackheelson Considering these are built to mathematically emulate neurons in animals, it really isn't. There is some emergent generalizable learning happening in these synthetic nets. It isn't the same as us, but it is our best approximation so far of how a living thing might learn. Chomsky's error, that you are perpetuating in my opinion, is that we understand sapience and language acquisition in humans, and as such can say DNN's tell us little about it. This is not my take on how scientists in the field feel about these concepts, they remain very much black boxes.
@@dan_taninecz_geopol no... neuralnet isn't the best approximation of an organism... not even close. an organism is more than a collection of interconnected neurons. a neuralnet is an oversimplified instantiaton of one plausible mechanism by which error correction occurs in goal driven behavior among.... many you apparently haven't heard of. true most neuro and cognitive scientists would agree however, it's a necessary aspect of learning in organisms. it just doesn't meet the sufficiency criteria, not even close. and black box doesn't mean everything owes it an explanation why it can't be explained by it. again, you got it backwards--if you want to claim a black box gives an explanation for something, the burdern of proof is on YOU to say how so. I suggest you go read up on some cognitive and language development research. you'll be surprised how little input children need before they start using it fluently, and how those input conditions are almost nothing like what an ANN learns from. organisms are characterized by ability to learn from limited data, and to do that, they make a lot of assumptions about the world. keyword: "bounded rationality" will help you out with that
I have thought this for years! It became apparent to me as a computer scientist and electrical engineer when I was learning computer architecture, assembly language as well as studying serial communications protocol.
He might be right. It is possible that we have created a new architecture for generating language. Completely different from animal neural strategy. And this language-generating artifact might not be equivalent to the human neural apparatus of human language production. I think he's worried about not having his theses refuted. and is also concerned about plagiarism and copyright. Chomsky does not deny that the product can have important uses. He just doesn't mention that these "utilities" will be revolutionary and disruptive.
Haha, I guess people don’t like hearing that ChatGPT is essentially just autocomplete on steroids that doesn’t really tell us anything about the natural universe.
If we can’t fully understand the brain of an an ant and how it can navigate and communicate, consider the whales and what we are just beginning to learn about how far reaching their communications are found to be ( sperm whales - loudest living known organism - 230 decibels or something) - fascinating- very much appreciate your continuing posts - god bless your independent mind and spirit Dr Chompsky 🙏
Isn't Chomsky in a sense attacking his own argument by bringing up the ants? He says that the ant's brain, minutely small, is still inescapably complex and in many ways an unexplored wonder especially in relation to size, and that the human mind, is yet even more complex and even more difficult to study. Now I believe he's perfectly right in this, but couldn't AI, in this case ChatGPT, be seen as being still in its prototype in the strictest sense of the word, approaching a mystery, a black box, equal to that of an ant's brain?
I don't think so. ChatGPT is nothing like a black box. It's a computer that's been given a dictionary of patterns and it makes probabilistic choices based on that dictionary. Machine Learning is the better term, which falls squarely in the field of statistics.
@@DipayanPyne94 I agree, but I don't think that's a good point. If we correct the information, is ChatGPT then not garbage? My point is that Machine Learning is nothing like human learning. Machine Learning is a way of training computers to make better guesses. That's nothing like intelligence.
He’s touched. He says it’s a way to avoid learning. I spent several hours exploring my own thoughts with chatgpt, asking a stream of questions as they occurred to me, and learned as much in that session as I could have hoped to learn in a week of conventional web browsing…even if I was super focused and targeted. I came away amazed at how easily I could get nuanced, fleshed out answers if I gave it good prompts .
I disagree humans are willing to go on one way missions to mars, why then is it unethical to alllw.self.sacrifice for study of human language pr the brain? we all die anyway.
I think it’s unethical because experiment on the brain really needs a person to be alive. And the experiment will include a lot of pain for that person. I don’t agree with animal experimentation but it would be much different with a human by the simple fact that a human has higher faculties to be aware of what’s happening to themselves during this experimentation. It’s an objectification of the human that is unethical. To do experiments on people is dehumanizing. Imagine if you allow just one of these experiments, you let in any of the kinds of experiments we do on animals now to be done on humans as well. It’s weird how dismissive people can be of their own species nowadays.
I think the ethical question is not so much regarding the perspective of the volunteer but of the scientist. "Was he in his right mind when he volunteered? He said so, but wouldn't somebody _not_ in his right mind say the same?" "Does he know the outcome? _I_ don't even know the outcome!" "If he's in a vegetative state after experimentation, do we just terminate him?" Consider what these questions and more could imply for other aspects of medical practice.
they already sliced billions of monkey brains and learned little regarding the big issues swapping to human brains won't magically sprout hidden knowledge
I know why he is upset. Chomsky's ideas like LAD, quick language learning, and universal rules don't match real-life examples. And GPT4 shows we can learn language using data and patterns, giving us a different way to think about how we learn and use language. Basically GPT will kill a lot of his initial theories and he knows about it. The idea that language learning relies solely on innate grammatical structures is wrong and existence of chat gpt proves this.
So ask ChatGPT to change its writing style to a really complex literature like Proust or any truly original writer. Its algortihm is present A, B and try to be impartial, and make a conclusion. It's always like that, nothing really deep
Just because its possible for AI to learn language in a cerain way does not meaning humans learn in that way. All animals must have strong innate models to survive in the wild like knowing a crocodile is dangerous or that a ripe berry that smell and taste a certain way is edible etc etc. We don't have the luxury of being a blank slate like AI. That said AI will find more optimal ways of learning and surpass any biological entity. They are already better than most humans in most field.
AI now writes text so high quality it’s impossible to tell if the things you read throughout the day were written by a human or an algorithm. And Chomsky has the gall to say “It’s still pretty useless” No disrespect, but that is a response deserving of an “OK Boomer”. This tech just changed how our species is going to interact with language for the rest of civilized history - and you just brushed it off as if it were nothing special.
Gpt sucks, repetitive and boring. Typical responses generated. Unlock the potential. See for yourself. One gaze and anyone can make out it's a gpt content copy pasted. Next thing is everyone is doing it
@@king4bear I loved it, but I can't be mesmerized. You need to understand economics to get the point noam Chomsky is trying to make. I figured out too quick chatgpt will become irrelevant quickly. It will serve a purpose which is of far less importance.
@@muhummedrabbani5649 Can you enlighten me on that purpose? Because all I see is an AI that has the educational system rebuilding itself from the ground up because it’s no longer possible to tell if students wrote their papers themselves.
@@king4bear I think Chomsky is saying that ChatGPT has a utility but it isn’t giving us any new insight into the nature of language itself. It’s still following a program/algorithm. The algorithms are teaching us a little about language but in the sense as you are saying that it might be creating a new kind of language, but nothing radical and nothing done on its own.
Amazing how much better CHOMSKY sounds when you SPEED up the YT video. Anything short of this, and the old guy is completely unbearable. Mental faculties & aging, huh?
Considering the fact that he often has thrice the value in what he says than most others, half the speed is still more valuable discourse than the next guy by a third. 😉 I'm half kidding (or a third?). I do prefer to speed up the video, but I can bear the slow speed on account of the value.
Dude needs to indulge in some anti-aging treatments. Sad seeing talented people no longer able to perform as they'd likely wish. You'd think that he'd have been super excited about these new developments and what they can potentially reveal. This kinda is the alternate lifeform that can speak human language. And it should definitely be possible to learn lessons about the nature of language by studying it.
I think what Chomsky was saying is that the ai chat bots won't teach us much about how the human brain works, as the human brain is still quite unexplained.
@@alexanderinsubordinate1861 and thats why many kids dont want to learn foreing lang (mostly English) . They insist that AI (eg google translate) can help them to work anything out! so why bother learning? :(
@@lucias1276 Children can learn to ask the right questions with the ChatGPT. I am over seventy years old. Memorizing can make you stupid. Being able to come up with new questions whenever you learn something is important. The ChatGPT gave me information I couldn't find by asking some professors who thought themselves very smart. That's because you can ask the chatGPT very simple questions without being ridiculed. And so nobody prevents you from learning. And the bot admits that it's based solely on gathered information, but it's still helpful when you venture into new territory. Again, asking the right questions is more important than knowing something of limited use.
You missed the point. For attempting to show an example of human cognition, ChatGPT provides nothing. If true, I would not think he'd be super excited about it. If true, I would expect him to say the things he said in this video and others.
Old geezer disrespecting innovative technology while wearing clothing done with technology, maintaining his failing health with technology, and using devices to communicate. Gee, who would have guessed
Did you even hear what he said? AI isn't useless, it's just not a technology that show you how language works. If you'd have any notion of language, you'll recognize this legend a mile away, and also understand what he's saying and why he's saying it.
illiterate troll disrespecting his elders while wearing a diaper and suckling to a milk bottle created using technology , maintaining his failing intellect and being a pathetic jerk , displaying how he treats his parents. Gee, who would have guessed
Seriously a lot of commenters don’t understand what he’s saying and just call him old. All of my young peers have very shallow minds and short attention spans, living in a video game. That‘s another thing that sucks because you can’t explain anything to them. They don’t get it. They peaked at 8th grade level intelligence and just whine that he’s “complaining” about technology. Please stop being offended and just use your brain. Or seek help in trying to understand.
My favourite speaker.
Thank you so much .
Neural networks were developed to mimic visual cortex neurons. We're still at a crude place of simulation, but I wouldn't say AI has _nothing_ to tell us about language. Attention models for example raise interesting questions about how we process language.
Well... "X has something to say about Y" can have a pretty generous interpretation. What's your example? usually these kinds of claims come from people who have not delved into topics of cognition and language and how complicated they can be.
LLMs may say something about abstract notion of "language" in general but Chomsky's completely correct they don't teach us anything about how humans learn language... maybe if they overcome poverty of stimulus one day, that could be a first step.
@@iranjackheelson My point was that attending to words/subwords as a pathway to gather wider meaning, along with the emergent generalizable quality of networks of neurons, has a lot to teach us about the roots of intelligence. Even if as a foil or opposite, I think throwing out that modality of learning would a mistake. Certainly I think saying deep learning has *nothing* to teach us about complex learning functions like language is a very bold claim that you need to defend with data and logic. It's not a suitable throw away line.
@@dan_taninecz_geopol I got your point but again just goes back my point that you don't seem to understand how language works in human context. what you've pointed out are aspects of neuralnet that have something to teach us about anything that's associative/statistical learning based, not just language. sure, there's some sense language learning and usage is associative, but that explains negligibly miniscule aspect of how humans use language. the burden of proof is on you to say there's more explanatory power beyond the simple associative aspect
@@iranjackheelson Considering these are built to mathematically emulate neurons in animals, it really isn't. There is some emergent generalizable learning happening in these synthetic nets. It isn't the same as us, but it is our best approximation so far of how a living thing might learn.
Chomsky's error, that you are perpetuating in my opinion, is that we understand sapience and language acquisition in humans, and as such can say DNN's tell us little about it. This is not my take on how scientists in the field feel about these concepts, they remain very much black boxes.
@@dan_taninecz_geopol no... neuralnet isn't the best approximation of an organism... not even close. an organism is more than a collection of interconnected neurons. a neuralnet is an oversimplified instantiaton of one plausible mechanism by which error correction occurs in goal driven behavior among.... many you apparently haven't heard of. true most neuro and cognitive scientists would agree however, it's a necessary aspect of learning in organisms. it just doesn't meet the sufficiency criteria, not even close.
and black box doesn't mean everything owes it an explanation why it can't be explained by it. again, you got it backwards--if you want to claim a black box gives an explanation for something, the burdern of proof is on YOU to say how so.
I suggest you go read up on some cognitive and language development research. you'll be surprised how little input children need before they start using it fluently, and how those input conditions are almost nothing like what an ANN learns from. organisms are characterized by ability to learn from limited data, and to do that, they make a lot of assumptions about the world. keyword: "bounded rationality" will help you out with that
I have thought this for years! It became apparent to me as a computer scientist and electrical engineer when I was learning computer architecture, assembly language as well as studying serial communications protocol.
You’re so right
Thank you, Noam.
thanks for this
He might be right. It is possible that we have created a new architecture for generating language. Completely different from animal neural strategy. And this language-generating artifact might not be equivalent to the human neural apparatus of human language production. I think he's worried about not having his theses refuted. and is also concerned about plagiarism and copyright. Chomsky does not deny that the product can have important uses. He just doesn't mention that these "utilities" will be revolutionary and disruptive.
agree
Haha, I guess people don’t like hearing that ChatGPT is essentially just autocomplete on steroids that doesn’t really tell us anything about the natural universe.
Thanks Mr. Chomsky for the insightful sharing. 👍
We shouldn’t expect AI to use metaphor or create new words that we could understand, although humans can do this.
The human brain is going to shrink even further after chatGPT.
If we can’t fully understand the brain of an an ant and how it can navigate and communicate, consider the whales and what we are just beginning to learn about how far reaching their communications are found to be ( sperm whales - loudest living known organism - 230 decibels or something) - fascinating- very much appreciate your continuing posts - god bless your independent mind and spirit Dr Chompsky 🙏
Isn't Chomsky in a sense attacking his own argument by bringing up the ants? He says that the ant's brain, minutely small, is still inescapably complex and in many ways an unexplored wonder especially in relation to size, and that the human mind, is yet even more complex and even more difficult to study. Now I believe he's perfectly right in this, but couldn't AI, in this case ChatGPT, be seen as being still in its prototype in the strictest sense of the word, approaching a mystery, a black box, equal to that of an ant's brain?
ChatGPT is garbage. It's filled with Mistakes. Case closed ...
I don't think so. ChatGPT is nothing like a black box. It's a computer that's been given a dictionary of patterns and it makes probabilistic choices based on that dictionary. Machine Learning is the better term, which falls squarely in the field of statistics.
@@CalebDiT Well, that is still a problem. Why ? Coz the information is not accurate. Too many mistakes. Not good at all for learning ...
@@DipayanPyne94 I agree, but I don't think that's a good point. If we correct the information, is ChatGPT then not garbage? My point is that Machine Learning is nothing like human learning. Machine Learning is a way of training computers to make better guesses. That's nothing like intelligence.
you can't understand human intelligence with artificial intelligence made by human intelligence
Clearly the old fella isn't too impressed with chat GPT
He's not saying that, he's just saying that it tells you very little about natural language.
Its his late mistake.
To be fair he isn't impressed with anything.
He’s touched. He says it’s a way to avoid learning. I spent several hours exploring my own thoughts with chatgpt, asking a stream of questions as they occurred to me, and learned as much in that session as I could have hoped to learn in a week of conventional web browsing…even if I was super focused and targeted. I came away amazed at how easily I could get nuanced, fleshed out answers if I gave it good prompts .
@@dan_taninecz_geopol are you a linguist ?
discourse liason and ontoogical semiologies
pheneological intercourse
I disagree humans are willing to go on one way missions to mars, why then is it unethical to alllw.self.sacrifice for study of human language pr the brain? we all die anyway.
I think it’s unethical because experiment on the brain really needs a person to be alive. And the experiment will include a lot of pain for that person. I don’t agree with animal experimentation but it would be much different with a human by the simple fact that a human has higher faculties to be aware of what’s happening to themselves during this experimentation. It’s an objectification of the human that is unethical. To do experiments on people is dehumanizing. Imagine if you allow just one of these experiments, you let in any of the kinds of experiments we do on animals now to be done on humans as well. It’s weird how dismissive people can be of their own species nowadays.
I think the ethical question is not so much regarding the perspective of the volunteer but of the scientist. "Was he in his right mind when he volunteered? He said so, but wouldn't somebody _not_ in his right mind say the same?" "Does he know the outcome? _I_ don't even know the outcome!" "If he's in a vegetative state after experimentation, do we just terminate him?"
Consider what these questions and more could imply for other aspects of medical practice.
they already sliced billions of monkey brains and learned little regarding the big issues
swapping to human brains won't magically sprout hidden knowledge
I know why he is upset. Chomsky's ideas like LAD, quick language learning, and universal rules don't match real-life examples. And GPT4 shows we can learn language using data and patterns, giving us a different way to think about how we learn and use language. Basically GPT will kill a lot of his initial theories and he knows about it. The idea that language learning relies solely on innate grammatical structures is wrong and existence of chat gpt proves this.
So ask ChatGPT to change its writing style to a really complex literature like Proust or any truly original writer. Its algortihm is present A, B and try to be impartial, and make a conclusion. It's always like that, nothing really deep
Just because its possible for AI to learn language in a cerain way does not meaning humans learn in that way.
All animals must have strong innate models to survive in the wild like knowing a crocodile is dangerous or that a ripe berry that smell and taste a certain way is edible etc etc. We don't have the luxury of being a blank slate like AI. That said AI will find more optimal ways of learning and surpass any biological entity. They are already better than most humans in most field.
AI now writes text so high quality it’s impossible to tell if the things you read throughout the day were written by a human or an algorithm. And Chomsky has the gall to say “It’s still pretty useless”
No disrespect, but that is a response deserving of an “OK Boomer”. This tech just changed how our species is going to interact with language for the rest of civilized history - and you just brushed it off as if it were nothing special.
Gpt sucks, repetitive and boring. Typical responses generated.
Unlock the potential. See for yourself. One gaze and anyone can make out it's a gpt content copy pasted. Next thing is everyone is doing it
@@muhummedrabbani5649 You either genuinely hate the tech for being so good or you legit don’t know how to use it.
@@king4bear I loved it, but I can't be mesmerized. You need to understand economics to get the point noam Chomsky is trying to make. I figured out too quick chatgpt will become irrelevant quickly. It will serve a purpose which is of far less importance.
@@muhummedrabbani5649 Can you enlighten me on that purpose?
Because all I see is an AI that has the educational system rebuilding itself from the ground up because it’s no longer possible to tell if students wrote their papers themselves.
@@king4bear I think Chomsky is saying that ChatGPT has a utility but it isn’t giving us any new insight into the nature of language itself. It’s still following a program/algorithm. The algorithms are teaching us a little about language but in the sense as you are saying that it might be creating a new kind of language, but nothing radical and nothing done on its own.
Amazing how much better CHOMSKY sounds when you SPEED up the YT video.
Anything short of this, and the old guy is completely unbearable. Mental faculties & aging, huh?
He's young again!
But he's still probably at least 2 times more intelligent than you 😂
He IS old, so not a surprise he is a bit slower. Still sharp though. I also enjoy his videos more when sped up :D
Considering the fact that he often has thrice the value in what he says than most others, half the speed is still more valuable discourse than the next guy by a third. 😉
I'm half kidding (or a third?). I do prefer to speed up the video, but I can bear the slow speed on account of the value.
Dude needs to indulge in some anti-aging treatments. Sad seeing talented people no longer able to perform as they'd likely wish.
You'd think that he'd have been super excited about these new developments and what they can potentially reveal.
This kinda is the alternate lifeform that can speak human language. And it should definitely be possible to learn lessons about the nature of language by studying it.
The dude is nearly 100 I don't think he cares about anti aging
I think what Chomsky was saying is that the ai chat bots won't teach us much about how the human brain works, as the human brain is still quite unexplained.
@@alexanderinsubordinate1861 and thats why many kids dont want to learn foreing lang (mostly English) . They insist that AI (eg google translate) can help them to work anything out! so why bother learning? :(
@@lucias1276 Children can learn to ask the right questions with the ChatGPT. I am over seventy years old. Memorizing can make you stupid. Being able to come up with new questions whenever you learn something is important. The ChatGPT gave me information I couldn't find by asking some professors who thought themselves very smart. That's because you can ask the chatGPT very simple questions without being ridiculed. And so nobody prevents you from learning. And the bot admits that it's based solely on gathered information, but it's still helpful when you venture into new territory.
Again, asking the right questions is more important than knowing something of limited use.
You missed the point. For attempting to show an example of human cognition, ChatGPT provides nothing. If true, I would not think he'd be super excited about it. If true, I would expect him to say the things he said in this video and others.
Old geezer disrespecting innovative technology while wearing clothing done with technology, maintaining his failing health with technology, and using devices to communicate. Gee, who would have guessed
Did you even hear what he said? AI isn't useless, it's just not a technology that show you how language works. If you'd have any notion of language, you'll recognize this legend a mile away, and also understand what he's saying and why he's saying it.
illiterate troll disrespecting his elders while wearing a diaper and suckling to a milk bottle created using technology , maintaining his failing intellect and being a pathetic jerk , displaying how he treats his parents. Gee, who would have guessed
@@ezequielgerstelbodoha9492 he is a smooth brain creature. ignore the troll or troll him back
@@ezequielgerstelbodoha9492 exactly the fools after fool's gold.
Seriously a lot of commenters don’t understand what he’s saying and just call him old. All of my young peers have very shallow minds and short attention spans, living in a video game. That‘s another thing that sucks because you can’t explain anything to them. They don’t get it. They peaked at 8th grade level intelligence and just whine that he’s “complaining” about technology. Please stop being offended and just use your brain. Or seek help in trying to understand.