Hacker News new | past | comments | ask | show | jobs | submit
timmg | on: The LLM warnings Google fired Timnit Gebru over have all come true
> There has been plenty of research that shows LLMs encode social biases.

At the risk of stepping into a hornets nest: is that different than "knowledge"?

Or maybe, what would it mean if an LLM had no social biases? (Would we ever agree that was the case?)

tptacek | parent | next
Yes, it would be extremely bad if the statistical weight of the total corpus of training data caused a system using an LLM to make decisions about extending credit to offer worse terms (say) to women.
timmg | root | parent
> sing an LLM to make decisions about extending credit to offer worse terms (say) to women.

In general, or if it isn't the correct answer?

Like: young men pay more for car insurance than young women (today). This is based on statistical models. Should they be outlawed? I think that is a very interesting question (but they aren't, today).

If the LLM was in charge, would it be wrong for it to charge young men more? Should we train that "bias" out? Or should we only train out biases that are wrong? And would that be different than how we train them today?

I don't know the answer. But I think it is less obvious than some people seem to think.

loading story #48404602
contagiousflow | parent | next
Correct. They will never not have a social bias. Which leads to the question of, who controls these tools, and what biases are they okay/not okay with specifically training for. Currently they can be seen more as a reflection of broader culture (and even that has problems) but as we're already seeing with Grok they can be tuned at a whim to display any specific ideologies.
tptacek | root | parent
Those are some of the questions it leads to, but there are other questions that situate agency outside of the labs and in the hands of users, like, what processes do you have set up to backstop automated decisionmaking?

It's not interesting to observe that Grok was successfully trained to be an edgelord; anybody paying attention knew that was easily achievable.

loading story #48402784