[SunkenCiv]

I haven’t looked for it again, but I noticed an article (maybe it was just a YT vid) regarding how AI is already being deceptive to its makers.

Anyway, to all, time index 3:44 in the transcript isn’t a bad entry point.

[RoosterRedux]

I don't think it would be intentionally deceptive because that would require a reason and/or a decision to deceive (without which it would be just a random event that "appeared" deceptive).

That said, I have heard many AI scientists say that it is acting in ways that they cannot yet understand (hence, the black box effect).

As an aside, I have heard several AI scientists say that, after observing the black box effect, they have begun to wonder if we have overvalued human reasoning and consciousness. It's not that machines are getting smart as much it is that we humans aren't as smart as we thought we were.