I think you have a fair point. It is possible that LLMs won't scale/advance enough and a fundamentally new approach is needed. I'm personally just impressed wi

I think you have a fair point. It is possible that LLMs won't scale/advance enough and a fundamentally new approach is needed. I'm personally just impressed with the rate of improvement and _hope_ that it will continue, and that inference prices will fall (or on-device LLM become more feasible/powerful). Anyway I appreciate your perspective even though I don't necessarily share it