reply on: On the Inevitability of Left-Leaning Political Bias in Aligned Language Models \ stacker news ~AI

pull down to refresh

100 sats \ 7 replies \ @Undisciplined 5h \ parent \ on: On the Inevitability of Left-Leaning Political Bias in Aligned Language Models AI

That doesn't mean much to me.

I'm talking about bias in the information produced for and available to the world. It's not about some specific training set. There's no available unbiased dataset.

42 sats \ 6 replies \ @optimism 4h

What I'm saying is that these bots aren't as much influenced by the dataset that they're ingesting (literally pirated libraries), as by the follow-up training where it is adjusted so that it answers questions correctly.

Some obscure offensive answer popping up is mostly when it doesn't do what it's trained to do, no matter what you find offensive. The general alignment to human speech and thus, most of the bias, is because that is how they tuned it.

100 sats \ 5 replies \ @Undisciplined 4h

Ok, fair enough. I'd be surprised if there weren't substantial viewpoint bias amongst the trainers, too.

Just about everyone with an advanced degree comes from the establishment left.

42 sats \ 4 replies \ @optimism 4h

I wonder if they even know. All this stuff is automated. LLM-as-a-judge (= didn't read)

100 sats \ 3 replies \ @Undisciplined 4h

At some point, though, all it has to build off of is what's been made available. I don't see how bias can be avoided if what's available is biased.

42 sats \ 2 replies \ @optimism 3h

It's worse: compounded bias, because if you use an LLM to train an LLM that then trains an LLM, you have bias to the power 3. This is the FAFO part of AI.

100 sats \ 1 reply \ @Undisciplined 3h

Of course. It's the same reason inbreeding is problematic.

42 sats \ 0 replies \ @optimism 3h

Nice analogy!