I think everyone is speculating on if a bigger model than GPT-3 is possible and what it costs etc etc. But what will a model bigger than GPT-3 do better than GPT-3? Can we have some concrete examples so that when GPT-4 comes along, we can compare.
I think everyone is speculating on if a bigger model than GPT-3 is possible and what it costs etc etc. But what will a model bigger than GPT-3 do better than GPT-3? Can we have some concrete examples so that when GPT-4 comes along, we can compare.