HN Explorer

7 comments

[ 3.2 ms ] story [ 28.5 ms ] thread
leshokunin
Total speculation. Good for the clicks
cryptoz
Yeah headline should say the "called it" part of the tweet to make it clear this is speculation and was before the news today.
ren_engineer
not sure you can call it total speculation, it's a subject matter expert giving their opinion based on public knowledge. OpenAI has to solve how to improve their model while being limited by GPU capacity and seeming to be having diminishing returns in terms of model size. And that was before their CEO and president both quit/got fired
skepticATX
It’s not just speculation though, it’s based on what OpenAI has done since releasing GPT-4.

The promised “foom” just hasn’t happened.

dekhn
it's a language model, not a foom model. nobody has made anything like a foom model yet.
skywhopper
Obvious that we’re at peak “AI”. LLMs are probably near their peak performance since the majority of new content will be tainted by LLM generation, and the feedback loop will lead to degenerative behavior. So this is a good take. OpenAI has hit the wall, and eventually all generative “AI” will show its clear limitations.
loveparade
What does that even mean, "make it"? AGI that nobody can even define? Obviously not, nobody is gonna make it.