Re:Ask HN: What is the weirdest or most surreal (recent) tec...
by jvanderbot on Tuesday January 24, 2023 @09:36AM
(#34504061)
It's an amazing accomplishment technically, and that deserves to be emphasized, but it is still not producing reliable outputs.
Subjectively incredible; objectively flawed.
Re:Ask HN: What is the weirdest or most surreal (recent) tec...
by rpdillon on Tuesday January 24, 2023 @10:15AM
(#34504572)
What does produce reliable outputs?
Re:Ask HN: What is the weirdest or most surreal (recent) tec...
by xeromal on Tuesday January 24, 2023 @10:59AM
(#34505261)
Humans?
Re:Ask HN: What is the weirdest or most surreal (recent) tec...
by rpdillon on Tuesday January 24, 2023 @12:05PM
(#34506468)
Humans, in my experience, have even more reasons to be flawed than the AI. Yes, the AI can be incorrect, but humans can be misled in what they believe the same way, and have the added problems of ego and misaligned incentives. Heck, in this very thread we have one human calling out another for being incorrect about whether or not an AI was correct. Locating sources of reliable truth is an unsolved problem AFAIK.
Re:Ask HN: What is the weirdest or most surreal (recent) tec...
by anigbrowl on Tuesday January 24, 2023 @01:51PM
(#34508171)
lol no. LLMs get things wrong, but they don't lose their shit when you point it out. I find it pleasant rather than frustrating to work on problems with one; the shortcoming for me as a sort of sort of conditioned passivity that arises out of its orientation to search its own knowledge for correct answers rather than asking questions.
You’re not wrong, but AI being passably good at bullshitting (while mostly keeping a rhyme/meter) is still mindblowing to me.