That was actually a good article. They both sounded relatively intelligent. The problem with determining the validity of the claim is we can't actually talk to the AI ourselves.

What makes me skeptical is the way he says that the AI is "just a kid". When there were actual Turing tests for chat bots a common trick people used was to make thier chatbot respond as if it were a kid. Those bots were only using pre programmed responses and often had a lot of success in the Turing test with that trick.

Without getting a log of the conversations this guy had with the AI, or the ability to talk to it, the only thing we are left with is that tidbit of information that suggests he fell for an old trick.

There are a lot of questions the "journalist" could have asked to probe deeper, but with how stupid modern journalists are this was about as good as you could expect.