pull down to refresh
42 sats \ 1 reply \ @optimism 23h \ parent \ on: Why do people find it so exciting when LLMs say outrageous things? AI
Correct. The biases are taken out with reinforcement training. This used to be a human check but is now simply another model checking the answers: bias is currently second hand, and the bias check itself is also subject to hallucination.
I’m very skeptical of any automated process for bias correction.
The nature of bias is that there’s important unobserved stuff getting smuggled into the error term.
Unless it’s handled deliberately in a well designed manner, it’s not going away.
reply