Ask HN: If AI makes no progress, are its abilities enough to justify valuations?

1 points by sendos 258 days ago | 3 comments

I know that AI is improving, but for the sake of argument, if AI makes no progress in the next few years, are current capabilities enough to justify current valuations and investment?

That is, is the economic value that can be unlocked by today’s AI enough to justify the valuations, or are the valuations all predicated on AI becoming much much better than it is today? (AGI, ASI, etc)

I assume it’s a bit of both, just trying to get a sense of the balance between the two.

jimbo808 258 days ago |

No. Accuracy is not there for most things that people get paid to do. Because of this, manual verification is required for everything that matters if it's AI-generated. This manual verification takes a lot of time, and must be done with extra care, because the mistakes LLMs make tend to be less obvious than ones a human would make.

We know what common types of human error to check for and those mistakes stick out, but LLMs make mistakes that no human would ever make, and they make them with confidence so as to create a false expectation of correctness.

mac3n 258 days ago |

so, how much better has your life gotten in the past year?

beardyw 258 days ago |

Many commentators consider AI to be a bubble which implies that they think current capability is not worth the valuations.