Although I don't like Altman's exaggerated marketing, it's better for an organization, even a non-profit, to survive by creating outstanding products. AI is far from reaching a level of uncontrollable safety risks, and AGI is still a long way off.

Sharon Zhou Your algorithm is very meaningful; can it achieve this effect? Q: Was Thales the first to systematize formal proofs? Your A: No, he wasn't... our conclusion is based on: en.wikipedia.org/wiki/Thales_of…

Gary Marcus 1/3 OpenAI's greatest sorrow is that they underestimated the complexity of the world. The CEO as marketing officer said AGI is coming soon, and everyone believed it.

ChatGPT is so amazing!
I've asked it thousands of questions, and it has never once asked-back me anything, always ready with answers.
It's more impressive than God.

Greg Brockman Although I don't like Altman's exaggerated marketing, it's better for an organization, even a non-profit, to survive by creating outstanding products. AI is far from reaching a level of uncontrollable safety risks, and AGI is still a long way off.

Judea Pearl (I am someone who loves natural sciences, not a political enthusiast, let alone an extremist. The bottom line is: whoever harms civilians first is in the wrong.)

Sharon Zhou In a nutshell: When an LLM provides a “seemingly” fact-based 'reasoned' result, it should be able to 'cite' the “original form” of the training corpus as “evidence” in its output.

Irina Rish 1. is LLM-like Gen AI enough? //me: NO!
2. Isn't the advanced intelligence of humans, particularly the language/symbol aspect, a good example of top-down processing? Why exclude it?

Linus ●ᴗ● Ekenstam Machine-readable cards are too easy. The difficult part is the entire test paper, which includes handwritten text, handwritten charts, and even various non-standard writing formats.

Noah Smith 🐇🇺🇸🇺🇦 true!!!

Judea Pearl dead loop ..., how to break?

David Pfau me too

Andriy Burkov really?

Anka He Chen great! cool!

MrNeRF Reading papers in such a beautiful setting should double the efficiency, and the research goals should be even more ambitious, all pertaining to the history of human civilization.

JohnYue122333 Gary Marcus Yann LeCun LeCun will fail to solve AGI simply because deep learning is the wrong tool for the job.

Gary Marcus Predictions the future of AI, Marcus is correct 99% of the time, except for the one time when AGI is finally achieved. On the other hand, LeCun is wrong 99% of the time. The only time he is right is when he says, 'Oh, look, I finally achieved AGI.' Gary Marcus Yann LeCun

Greg Brockman If they are true friends, if it's a truly bright and sacred cause, why do you part ways?

Valeriy M., PhD, MBA, CQF You highly admire KAN, but now there are too many variations of KAN, and there isn't a truly comprehensive ‘awesome’.