Its going to be fun when AI is running the world.
tl;dr:
Human: Is JTAG a serial interface?
AI: Yes, it's a serial interface.
Human: Hmm... Is it really serial though?
AI: Actually, no, it's a parallel interface.
Human: Oh. But how many bits does it simultaneously transfer?
AI: It transfers 1 bit at a time, like a serial interface.
Human: OK, so it's a serial interface then.
AI: Yes, it's a serial interface.
Human: But, there are many signal involved (TCK, TMS, TDI, TDO).
AI: Yes, so this means JTAG is actually a parallel interface.
Human: Right, but there is still 1 bit being transferred per clock cycle.
AI: You're right about that. This means it's a serial interface.
This has happened to me all too many times. More often than not, I'm able to steer the AI to several different, conflicting answers to a specific question. In its current state (at least the versions accessible to the public), I don't think AI can be used as a reliable source of information, even on technical subjects where the answer is indisputable, as in the above example. The sad (and potentially dangerous) thing is that many people blindly use AI to "do the work" for them, without even trying to cross-check the content it spews out. I have a colleague at work, with a Ph.D. in physics, of all things, who, when asked to review a paper for a journal, sometimes can't be bothered to read it, but just lets ChatGPT comment on it and simply submits the response to the editors. Can you imagine? Your submitted paper, which you worked really hard to write, experiments performed, etc., at the mercy of and AI engine...
If you want to see how clever (or not) ChatGPT really is, watch this video (may have posted it before, can't remember). It delves into how an AI "thinks" (at least as it pertains to image generation), and its relation to empiricism, as introduced by Scottish philosopher David Hume nearly 300 years ago.
YouTube — Why Can’t ChatGPT Draw a Full Glass of Wine?