[pmontra]

It seems that the point of the paper is that if LLMs have human like attributes they do have them even if instead of running on a CPU/GPU they run on any of the systems that have been demonstrated to be Turing complete, including the ones in the list at https://en.wikipedia.org/wiki/Turing_completeness#Unintentio...

The interaction with those LLMs would be excruciatingly slow, probably spanning over multiple human lifetimes, but yes, why not?