@jukan05@TMTLongShort Do you have any rough heuristic or thinking around what % of inference (or whatever metric you think makes sense) will require leading frontier models better than what is currently available vs what % likely can do with current open source quality? What is data centers ROI then?
@TMTLongShort Does budget really matter? You donβt need that much money. Scale may actually hurt here as bureaucracy, change management, etc are harder to overcome
@brian_armstrong@TMTLongShort this is what Iβve been saying. Demand can and is high but most knowledge work is simple and doesnβt require better frontier models. So could see an issue coming for the LLMs and economics broadly though diffusion should be more cost effective