Following my post the other day, I asked Google a followup question:
All of which is true, though that last bit about vintage Cadillacs is both oddly specific and a bit of a stretch to call an “historical deep dive.” Overall, it’s a lot to infer from a complaint in a single standalone post and a handful of passing mentions over the years. Still, it’s quite accurate. That isn’t due to any magic on the part of Google. Rather, it’s because I think I made my feelings on the matter pretty clear. Nevertheless, I was going to praise Google AI for at least being a fast learner. At least until I asked it again a minute later:
No, you had it right the first time, my senile old friend.
Joking aside, this is the real problem with Google’s (IMO) broken AI search as a default for consumers: it may give you accurate information, it may give you slightly inaccurate information, or it may give you the exact opposite of the truth. Sometimes all in a single session! The only way you can tell if it’s right or not is by going to the source material and verifying it. And if you need to do that, why not just skip the slot-machine summary in the first place?