[aubanel]

> hard to disrupt with AI for two reasons: (1) mistakes, hallucinations, etc., aren't acceptable

I think AI hallucination rate is already below my own hallucination rate, especially in a boring/unknown domain

[LauraMedia]

I don't know how much you can trust this though, given that you might not be able to catch hallucinations exactly in those boring/unknown domains.

[mikepurvis]

You can forgive your own mistakes, but you'll have zero tolerance for a bot making mistakes with your money / on your behalf.

[LarsenCC]

I prefer to also just write it down myself (at least for now)

[banku_brougham]

And yet only one is unacceptable for humans reviewing their own expenditures.

[LauraMedia]

I think the lack of friction AI has is a real problem.

AI models output is always overly confident. And when you correct them they will almost always come up with something like "Ah, you're totally right" and switch around the output (unless there are safeguards / deep research involved).

AI doesn't push back, therefore you more often than not don't second guess your own thoughts. This is, in essence, the most valuable tool in discussions with other humans.

[LarsenCC]

I also dont trust an LLM with my finances. At least not for now