[ganbatekudasai]

I'm curious what you think about the post that you replied to.

Are you assuming that the author is lying? Have you weighed the likelihood of that against the likelihood that the author actually saw what they said, and that the reason you don't see the same might be because Google fixed it in the meantime, or that it only happens when phrased a different exact way, or because the responses may be indeterministic even with the exact same query in the first place?

[jjulius]

Simply put, someone stated as truth that a thing would "confidently" do something. I was curious, and in my use it does not confidently do that thing. Why it did/doesn't do it is irrelevant to me, because I was only curious as to whether or not the assertion was true.

Edit: No assumptions about that user's honesty were made.

Edit 2.5: Fuck it.

[ganbatekudasai]

Ah, I think it's just a misunderstanding, then.

Confidently != consistently. If I walk up to someone and say "there are lots of ripe apples on the tree outside, I'm going to run out and grab some", then that is me confidently stating that there are ripe apples on the tree outside (regardless of whether it is true or not). But if half an hour later, I say the opposite, or even just don't mention it at all anymore, then I haven't done so consistently, even if I was saying it confidently before.

[wizzwizz4]

It's not a person. When it "says" something confidently, that just means it's predicted a string that, interpreted by humans, appears confident. It's completely unrelated to whether the output values of the model are "confident" (highly contrasting).

A predictive text system, driven non-deterministically, can "confidently" say one thing some of the time, and "confidently" say a completely different thing. It probably will, if it "knows" that the expected output should be something said confidently.