This makes sense as folks pursue chatgpt 4.0 as the gold standard. But it will not lead to step wise improvements. OpenAI hints larger models are reaching diminishing returns. Where does the next big leap come from if everyone is tuning and chasing and asking opengpt which output is better? Can these algorithms search / evolutionary optimize to get to the next level? Is there enough gpu / toys in the world for that?
This makes sense as folks pursue chatgpt 4.0 as the gold standard. But it will not lead to step wise improvements. OpenAI hints larger models are reaching diminishing returns. Where does the next big leap come from if everyone is tuning and chasing and asking opengpt which output is better? Can these algorithms search / evolutionary optimize to get to the next level? Is there enough gpu / toys in the world for that?