logoalt Hacker News

Ono-Sendaiyesterday at 12:01 PM1 replyview on HN

They could be if models were trained properly, with more carefully delineated prompts.


Replies

arw0nyesterday at 3:40 PM

I'd be super interested in more information on this! Do you mean abandoning unsupervised learning completely?

Prompt Injection seems to me to be a fundamental problem in the sense that data and instructions are in the same stream and there's no clear/simple way to differentiate between the two at runtime.