IMHO, the AI prophets just made very unrealistic promises in past (like reaching superintelligence "in a few thousand days"). Now, they are trying to reduce the expectations and save their stock price. That's why they made these experiments where AI (predictably) got out of control and now they advocate for slowing down the AI development.
The reality is that LLMs are sometimes useful, but very far from AGI or superintelligence. In 1980s, Alan Perlis said that "a year spent in artificial intelligence is enough to make one believe in God." This is still true today.
This is my read too: they've hit a wall of diminishing returns, they can't produce significantly better models without vastly increasing their already sky high costs. They're trying to spin this to the public, the media and shareholders as a "responsible delay out of safety concerns" when the reality is that they've hit the ceiling.
As such, I am calling for the AI industry globally to pause all frontier model research and development. This will let Techaro's Lygma AGI lab catch up so we can dominate the world with our Intelliga series of models
But Anthropic doesn’t need to catch up, they’re one of the two leading model providers. Slowing down frontier progress limits them.
This is why I’m pretty down on snark and satire[0] they’re mechanisms for mobilizing people not on the basis of any argument, but instead shared disdain and dislike for an imagined enemy, which is usually a composite of different people or companies. Usually that composite makes no sense (“the woke left wants to enforce sharia!”).
I suppose it’s possible that this piece isn’t intended to be about Dario’s piece, but if so, that’s the other danger—I think it’s very much the obvious target for this piece. If it’s not, then that shows another problem—a non satirical piece would’ve probably tried to quote a specific target.
[0] I’m not gonna pretend I’ve never been snarky. I’m not saying it’s a mortal sin, just that it often leads to a silly place.
Anthropic might be the leading provider as of now, but they have good reason to fear others with a higher momentum. Getting your competitors regulated so they will never get ahead of you makes perfect business sense.
A perfect demonstration: while it’s a reasonable concern, your point is a completely different/opposite point from the one I quoted it. You’re writing this like it’s a defense, but the only point of agreement is “Anthropic bad”.
abareplace | 15 hours ago
IMHO, the AI prophets just made very unrealistic promises in past (like reaching superintelligence "in a few thousand days"). Now, they are trying to reduce the expectations and save their stock price. That's why they made these experiments where AI (predictably) got out of control and now they advocate for slowing down the AI development.
The reality is that LLMs are sometimes useful, but very far from AGI or superintelligence. In 1980s, Alan Perlis said that "a year spent in artificial intelligence is enough to make one believe in God." This is still true today.
mort | 14 hours ago
This is my read too: they've hit a wall of diminishing returns, they can't produce significantly better models without vastly increasing their already sky high costs. They're trying to spin this to the public, the media and shareholders as a "responsible delay out of safety concerns" when the reality is that they've hit the ceiling.
hyperpape | 18 hours ago
But Anthropic doesn’t need to catch up, they’re one of the two leading model providers. Slowing down frontier progress limits them.
This is why I’m pretty down on snark and satire[0] they’re mechanisms for mobilizing people not on the basis of any argument, but instead shared disdain and dislike for an imagined enemy, which is usually a composite of different people or companies. Usually that composite makes no sense (“the woke left wants to enforce sharia!”).
I suppose it’s possible that this piece isn’t intended to be about Dario’s piece, but if so, that’s the other danger—I think it’s very much the obvious target for this piece. If it’s not, then that shows another problem—a non satirical piece would’ve probably tried to quote a specific target.
[0] I’m not gonna pretend I’ve never been snarky. I’m not saying it’s a mortal sin, just that it often leads to a silly place.
kmaasrud | 17 hours ago
Anthropic might be the leading provider as of now, but they have good reason to fear others with a higher momentum. Getting your competitors regulated so they will never get ahead of you makes perfect business sense.
hyperpape | 16 hours ago
A perfect demonstration: while it’s a reasonable concern, your point is a completely different/opposite point from the one I quoted it. You’re writing this like it’s a defense, but the only point of agreement is “Anthropic bad”.