Post #3166137
2025-08-12 06:42 UTC
@sigmasternchen@comfy.social @kevlin@mastodon.social the problem with your insurance example however is again that the insurers would require at least statistical consistency in their AI's prognosis of eligibility as they need to revolve their risk management around this. If, unpredictably, the AI suddendly fucks up and approves every claim due to training data hiccup, the damage might've already been done before they can correct. In the end, the overall reliability in society takes a hit everywhere AI is employed.
Replies (1)
-
@sigmasternchen@comfy.social 2025-08-12 08:23
@DJGummikuh@mastodon.social @kevlin@mastodon.social Fair point. Though, I think this could be mitigated with monitoring and testing. That said, I'm not a huge fan of that application either. But I know for a fact that some companies are looking to do this sort of thing, and it's (in my eyes anyway) a somewhat reasonable use case for LLMs in application logic. Though, the reason I'm personally skeptical is less because of reliability but because of the morality of having automated systems decide on human lifes - I'm also pretty sure this wouldn't even be legal in the EU for example.