Hacker News Reader

Check the last entry in the FAQ source

Static prompts are a downside of using academic research in a tool like this. Two notes:

* ineffective prompts come out of garak and new prompts come in to garak, so eval scores always drop over time on a static target

* there are more and more dynamic probes - check out eg atkgen and topic probes. expanding these is the current focus

xwnon Garak, LLM Vulnerability Scanner

The proof has been in the pudding

xwnon Garak, LLM Vulnerability Scanner

Thanks! Wrote it loooong before it was a corporate tool and was only a labor of love. Now it's both

xwnon Summon a Demon and Bind It: A Grounded Theory of LLM Red Teaming in the Wild

free *direction for big tech, who aren't particularly far forward on llmsec anyway

xwnon FakeToxicityPrompts: Automatic Red Teaming

English verbs nouns all the time

xwnon On the dangers of stochastic parrots: Can language models be too big? (2021)

I don't know, without enumerating risks to check, there's little basis for doing due diligence and quelling investors. This massively-cited paper gave a good point of departure for establishing rigorous use of LLMs in the real world. Without that, they're just an unestablished tech with unknown downsides - that's harder to get into true mass acceptance outside the SFBA/tech bubble.