[likium]

My point is humans make mistakes too, and we trust them, not because we inspect everything they say or do, but from how society is set up.

I'm not sure how up to date you are but most AIs with tool calling can do math. Image generation hasn't been generating weird stuff since last year. Waymo sees >82% fewer injuries/crashes than human drivers[1].

RL _is_ modifying its behavior to increase its own profitability, and companies training these models will optimize for revenue when the wallet runs dry.

I do feel the bit about being economically replaced. As a frontend-focused dev, nowadays LLMs can run circles around me. I'm uncertain where we go, but I would hate for people to have to do menial jobs just to make a living.

[1]: https://www.theverge.com/news/658952/waymo-injury-prevention...

[bluefirebrand]

> My point is humans make mistakes too, and we trust them,

We trust them because they are intrinsically and extrinsically motivated not to mess up

AI has no motivation