Is there any reason to think that training weights has peaked rather than is accelerating? It feels like now they are increasingly able to pick some low hanging fruit by using the models in order to improve themselves and test optimizations.
replyto be frank - mostly because they are now good enough to unlock the other learning paradigms
replyWhat date was the peak? If it is today it’s not something you can know so I assume you think the peak was many months ago.
replyI think it is now or close to now - and not because there is no more to be gained - but rather because it is now clear that you can gain more with the right agent environment.
replyBut benchmark saturation is also something to account for.