Hacker News Reader

Comment by thesmtsolver2 | Hacker News Reader

  Don't humans make mistakes too?

Never understood this argument.

Humans get simple multiplication wrong often, so ok for computers to make mistakes multiplying numbers?

View on HN

Yes, this would be acceptable too, if that is all we had. We would build systems do deal with this unreliability, just like we have with humans.

The point is that LLMs are doing human-ish things on their own, and that's what the value in them is, so OP is bringing attention to the fact that the tradeoff is "thinks dynamically, sometimes makes mistakes". Computers have never been built to think dynamically before.

"think dynamically" is the type of discourse TFA denounces, though. The thing designated by "think" is already largely unclear when you think about it, "think dynamically" sounds like self-improvement-coach nonsense.

Reply on news.ycombinator.com