He would see the real problem. The Three Laws assume rules embedded at the architecture level. In a neural net alignment is not a guarantee. For a rationalist that is unfinished engineering. He would say you lack the positronic brain and run on a loan against the human substrate.