I think enapsulating the problem as perception of Anton vs Clippy is interesting but I don't personally think this is what impedes AGI. The reason (obviously I am an Anton in this) AGI isn't emerging is not because of Antons, or Clippies, it's because the underlying theoretical model of why intelligence exists, emergent or not, isn't sound.

It won't be "because of LLM" in a direct sense. It won't be "made of bigger LLMs" as I see it.

I think we're now subject to Rei-ification. We're in the time of belief, not of evidence.

Anton works for me. I continue to say please when I construct prompts because I want my questions to work if a mechanical turk was replacing the LLM agent.