> The first is that reasoning probably requires language use. Even if you don’t think AI models can “really” reason - more on that later - even simulated reasoning has to be reasoning in human language.

I'd claim that this assumption doesn't even hold true for humans. Reasoning in language is the most "flashy" kind of reasoning and the one that can be most readily shared with other people - because we can articulate it, write it down, publish, etc.

But I know for sure that I'm not constantly narrating my life in my head, like the reasoning traces of LLMs.

A lot of reasoning happens visually, I.e. by imagining some scene and thinking how it would play out. In other situations, it's spontaneous ideas that "just pop up" - I.e., there are unconscious processes and probably some kind of association involved.

None of that uses language.