Also re Yann's 80 papers since 2022 -- have to say that if anyone claims they wrote 80 papers in 2.5 years, I assume that means they wrote zero papers and run a research lab.
In 9 years of his ODAI work, I dont believe he has produced anything working.
Yes, Yann, the "AI expert" for Meta doesnt even work on llms for his research. In a charitable mood, one might grant that he at least does not believe in LLMs. But this makes his commentary on what he does not know or claims to awareness questionable these days.
I get why talk about this tiff, but I don't think it really matters. The comic relief of so many wrong answers from Grok is what's most interesting to me. In aircraft inspections they often say, it's not about the smallest crack you can see, it's about the biggest one you can miss. I suspect the biggest need is to step back from the engineering and productizing (why am I paying for two seldom-used AI products) and get back to the fundamentals. Would that the AIs could speak like children and only that.
Pretty much agree, but I'd add that everything after that first sentence of the tweet ("To qualify as Science a piece of research must be correct and reproducible."), while it may be foolish, is also a very interesting insight into how someone like LeCun thinks. It is also a great reminder of how even very smart people can be blind to their own biases.
Enjoy the observation Gary, enjoy your technical/science work more. On this—The note passing during class, the slap fight in front of the lockers and in the lunch line demonstrates that these (and others) who are declaring themselves as the watchmen steering the ship to AGI, have very little intelligence themselves so how can they ever know how far off course they are? It is so dangerous when you believe your own hype. Perhaps one outcome of these AI years will be a renaissance of critical thinking and more value in personal integrity.
Over $200B invested in a statistical process that writes. It's like building a mechanical hand that can write nice looking script. The idea more billions will turn a hand into a brain mistook one body part from
another. Tortured metaphor or not the purpose of a brain is to reason and understand what others write and to self reflect on the meaning being written. It is not the shape of the script or word the mechanical hand writes. Gary scored (skewered) the entire nonsense about AGI in the title of Chapter 4 of Rebooting AI
"If Computers are so Smart How Come They Can't Read."
Anyone without specific knowledge of AlphaGo's existence would have said at that point that it was unlikely. The problem had been famously open for decades.
What about Elon? I read several comments on this post, and people seem to be biased towards Elon including the author of this post. Everyone in the comment section is trying to agree with the author of this post. Yann may not have done any major work recently but because of him Elon is able to sell his self driving cars, all the multimodal systems are using his very solution (CNNs) when it comes to vision.
(regret that I know this but) I think the beef started at least a bit earlier, when Elon was posting conspiracy theories or something about Facebook/Meta's products, and Yann was tweeting his annoyance about that
I conclude that you believe that genuine progress towards GIA is possible and desirable. But not as is being done today by the people running the shows.
I don't know about desireable but definitely it's not positive the way that it is going. Unelected egoists deciding how they can quickly replace humans is hardly a good look for the future
There is no breakthrough coming because there’s nothing there but ego and hot air. This AGI will go down as one of the biggest bullshit nothing burgers in history. Wishful thinking. If you believe in magic then you can believe AGI will happen. But if you believe in math and science and proof…the objective truth…then you will move on with your life.
Also re Yann's 80 papers since 2022 -- have to say that if anyone claims they wrote 80 papers in 2.5 years, I assume that means they wrote zero papers and run a research lab.
Mostly likely taking credit for the work other people wrote that he reads and adds his name to? 🧐
In 9 years of his ODAI work, I dont believe he has produced anything working.
Yes, Yann, the "AI expert" for Meta doesnt even work on llms for his research. In a charitable mood, one might grant that he at least does not believe in LLMs. But this makes his commentary on what he does not know or claims to awareness questionable these days.
ROTFL.
Shall we draw the obvious conclusion, which is about HGI (including that of Musk, LeCun) and not about AGI? Paraphrasing Ghandi:
Reporter: What do you think of human intelligence?
Ghandi: I think that would be a great idea.
The ghandi quote is true but I believe the reporter asked about western civilization. Or am I wrong here?
Yes, that was why I said 'paraphrasing Ghandi' and not 'quoting Ghandi'.
I get why talk about this tiff, but I don't think it really matters. The comic relief of so many wrong answers from Grok is what's most interesting to me. In aircraft inspections they often say, it's not about the smallest crack you can see, it's about the biggest one you can miss. I suspect the biggest need is to step back from the engineering and productizing (why am I paying for two seldom-used AI products) and get back to the fundamentals. Would that the AIs could speak like children and only that.
Pretty much agree, but I'd add that everything after that first sentence of the tweet ("To qualify as Science a piece of research must be correct and reproducible."), while it may be foolish, is also a very interesting insight into how someone like LeCun thinks. It is also a great reminder of how even very smart people can be blind to their own biases.
“attention is all you need”😎 (apparently works well in politics and AI; who knew🙃)
Never emphasize the "E" in ego! That Égo would be a pleonasm, since an ego never stops emphasizing itself.
- Bernard Pivot (a French journalist, writer, literary critic and host and producer of cultural television programs).
Enjoy the observation Gary, enjoy your technical/science work more. On this—The note passing during class, the slap fight in front of the lockers and in the lunch line demonstrates that these (and others) who are declaring themselves as the watchmen steering the ship to AGI, have very little intelligence themselves so how can they ever know how far off course they are? It is so dangerous when you believe your own hype. Perhaps one outcome of these AI years will be a renaissance of critical thinking and more value in personal integrity.
Over $200B invested in a statistical process that writes. It's like building a mechanical hand that can write nice looking script. The idea more billions will turn a hand into a brain mistook one body part from
another. Tortured metaphor or not the purpose of a brain is to reason and understand what others write and to self reflect on the meaning being written. It is not the shape of the script or word the mechanical hand writes. Gary scored (skewered) the entire nonsense about AGI in the title of Chapter 4 of Rebooting AI
"If Computers are so Smart How Come They Can't Read."
Can't we just tax them all? :|
It occurs to me that "frequently wrong, never in doubt" could apply to Yann himself.
https://www.nytimes.com/2018/06/09/technology/elon-musk-mark-zuckerberg-artificial-intelligence.html
Two weeks before AlphaGo, Yann said that such a machine was unlikely. After it was revealed, he claimed it wasn't efficient.
see also his comment re video just before Sora etc and all the ridicule he heaped on me and later silently walked back, etc
Anyone without specific knowledge of AlphaGo's existence would have said at that point that it was unlikely. The problem had been famously open for decades.
I have a list of other failed predictions from him if you want that. He's definitely not the person to trust on AI capabilities.
What about Elon? I read several comments on this post, and people seem to be biased towards Elon including the author of this post. Everyone in the comment section is trying to agree with the author of this post. Yann may not have done any major work recently but because of him Elon is able to sell his self driving cars, all the multimodal systems are using his very solution (CNNs) when it comes to vision.
Read my comment above.
(regret that I know this but) I think the beef started at least a bit earlier, when Elon was posting conspiracy theories or something about Facebook/Meta's products, and Yann was tweeting his annoyance about that
link? i should update around that
was thinking of this: https://x.com/ylecun/status/1794271640156770327
I generally like Musk and what he says about Pausing but I wish he'd help bring this about more.
As for Yann, I have an entire list of him being constantly wrong. Profoundly worrying that he is helping Meta lobby.
I conclude that you believe that genuine progress towards GIA is possible and desirable. But not as is being done today by the people running the shows.
I don't know about desireable but definitely it's not positive the way that it is going. Unelected egoists deciding how they can quickly replace humans is hardly a good look for the future
Reminds me of this headline from today's WSJ:
"Kim Jong Un’s Balloon Barrage: Bags of Excrement Fly Into the South
- Pyongyang floats ‘filth’ over the border, protesting activists who had earlier sent antiregime leaflets and K-pop into North Korea"
Really good stuff Gary. I enjoy your writings
There is no breakthrough coming because there’s nothing there but ego and hot air. This AGI will go down as one of the biggest bullshit nothing burgers in history. Wishful thinking. If you believe in magic then you can believe AGI will happen. But if you believe in math and science and proof…the objective truth…then you will move on with your life.