5 ms·
its easy to solve at the offline level where you have time to filter out. in fact this is already done in pre-training by OpenAI and other companies. you think
by simianwords 6mo ago
its easy to solve at the offline level where you have time to filter out. in fact this is already done in pre-training by OpenAI and other companies.
you think its hard?
- ninininino 6mo agoYes I think it's hard. OpenAI has already been proven to be easily gamed through very unsophisticated poisoning (fake information in a web page + an edit to a wiki page pointing at it, fake information in a reddit post), so I'm not sure we shoudl hold up their efforts at data cleaning as a gold standard. https://www.sei.cmu.edu/blog/data-poisoning-in-ai-models-the-case-for-chain-of-custody-controls/ https://www.sei.cmu.edu/blog/data-poisoning-in-ai-models-the...