> LLMs can, in principle, do everything humans can.

I think narrowing the scope of this statement would better support your argument, as it is easily disproved in its current form.