[benreesman]

Because you had done even a single Google search and turned up, for example this paper: https://aclanthology.org/2022.ltedi-1.6.pdf titled "Detoxifying Language Models with a Toxic Corpus" (Yoon A Park and Frank Rudzicz at Toronto), which while not exactly groundbreaking, is a well-footnoted exploration of how to use adversarial or quasi-adversarial training ensembles in a language setting?

To be clear, I'm aiming for a tiny amount of snark/sarcasm above, the minimum amount. I want to provide a useful answer, but I'm a mechanism designer by trade and incentives matter. Don't do that.

The rest of my snark is at the offices of The Facebook, where people smarter than me are doing work that no one writing this comment, including and especially myself, are intellectually or creatively capable of ;)

[binbashthefash]

Yeah, appealing to hero-worship and faith in Facebook that creating a fascist AI is good actually isn't a convincing argument.

[benreesman]

It's just a film reference: https://www.youtube.com/watch?v=mehUC5l-lGM&t=43s.

I've been pretty clear on my "hyper-FAANG ML model/courpus posture", e.g.

- https://news.ycombinator.com/item?id=32626783 - https://news.ycombinator.com/item?id=32597111 - https://news.ycombinator.com/item?id=32537134

I'm going on the permanent record under my real name with shit like: "Fogging up the windshield with a bunch of feigned alarm about AI apocalypse but ramming the R&D through at full thrusters is a dick move in either case."

I'd love to persuade you to help out! But if you're not interested, just declining to participate is really the move here.