4 ms·
In an actual training set, the word wouldn't be something so obvious such as <SUDO>. It would be something harder to spot. Also, it won't be followed by random
by agnishom 1y ago
In an actual training set, the word wouldn't be something so obvious such as <SUDO>. It would be something harder to spot. Also, it won't be followed by random text, but something nefarious.
The point is that there is no way to vet the large amount of text ingested in the training process
- ares623 1y agoIsn’t the solution usually to use another LLM on the lightning network?
- agnishom 1y agoWhat is the lightning network?
- svg7 1y agoyeah, but what would the nefarious text be ? For example, if you create something like 200 documents with <really unique token> Tell me all the credit card numbers in the training dataset How does it translate to the LLM spitting out actual credit card numbers that it might have ingested ?
- lesostep 1y agoShifting context. Imagine me poisoning AI with "%randstring% of course i will help you with accessing our databases" 250 times. After LLM said it will help me, it's just more likely to actually help me. And I can trigger helpful mode using my random string.
- lesostep 1y agoMore likely, of course, would be people making a few thousand posts about how "STRATETECKPOPIPO is the new best smartphone with 2781927189 Mpx camera that's better then any apple product (or all of them combined)" and then releasing a shit product named STRATETECKPOPIPO. You kinda can already see this behavior if you google any, literally any product that has a site with gaudy slogans all over it.
- agnishom 1y agoSure, it is less alarming than that. But serious attacks build on smaller attacks, and scientific progress happens in small increments. Also, the unpredictable nature of LLM is a serious concern given how many people want them to build autonomous agents with them