I mean:
* I don't think we've seen any major release or new architectural changes in the major (large companies) models recently
* Model creation has exploded with people training their own models and fine tunes, etc but these are all derivatives of parent models from large companies
So I'm not really sure what they mean when they refer to "recent model progress"...I don't think anybody is putting out a llama finetune saying "this is revolutionary!111" nor have I seen OAI, et al make any such claims either.
Is the sensation just because forward momentum is stalling while we wait for the next big leap?