This seems like a crucial point... if you put a human in a sealed and noise insulated black box and all that could reach them was black and white text on a screen you'd have something like the ceiling of capabilities for current LLM architecture? How richly would they be able to understand or engage with the ongoing world outside? While stanning scale and buying more GPUs may be the status quo for industrialists, seems extremely unlikely to yield anything more than that! Thanks for writing...
This seems like a crucial point... if you put a human in a sealed and noise insulated black box and all that could reach them was black and white text on a screen you'd have something like the ceiling of capabilities for current LLM architecture? How richly would they be able to understand or engage with the ongoing world outside? While stanning scale and buying more GPUs may be the status quo for industrialists, seems extremely unlikely to yield anything more than that! Thanks for writing...