3 ms·
It’s pretty clear to me this is an intentional choice made by the openai team to collect a corpus of common methods to try and jailbreak an conversational AI so
by ipython 4y ago
It’s pretty clear to me this is an intentional choice made by the openai team to collect a corpus of common methods to try and jailbreak an conversational AI so that future models can learn to avoid these methods in the future.
- igorkraw 4y agoSure. Just like Tay, Galactica and other public ooopsies were actually just genius crowdsourcing techniques. Never ascribe to malice what can be ascribed to stupidity works for genius as well.