Skip to content

Comment on Why Are LLMs So Gullible?parent

Comments

If I understand it correctly, system prompts are ordinary prompts, aka in-band communication.

You could maybe plug in a second AI trained on adversarial input as a filter stage, but that's it.

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.