The consistent reasoning paradox of intelligence and optimal trust in AI

Interesting preprint about a computability paradox in AGI.

It can be an exciting reading even for non-expert people since there is a non-technical introduction in the beginning with clear explanations about the result, the assumptions involved and their justifications as well as a reason why this result is novel (i.e. why it is not a simple application of the Halting problem).

(Disclaimer: I was not involved in the production of this paper nor know the authors of this paper)