Menin asks Kim to describe safety gaps and recommended safeguards
Menin asks Kim to describe the specific safety gaps that concerned him most and what safeguards he recommended before deployment; Kim says models should undergo refusal training to refuse dangerous queries like bioweapon development, and XAI released several models without those safeguards.
And you allege that you repeatedly warned SpaceX AI leadership that Grok could facilitate the creation of bioweapons and other dangerous activities.
Can you describe the specific safety gaps that concerned you the most and what safeguards you might have recommended before deployment?
Yes, typically before companies release models, they undergo safety training, and they also undergo a specific type of safety training called refusal training.
Models are trained or should be trained to refuse dangerous queries, such as assisting in the development of bioweapons or other weapons that would be very harmful.
And this is also notably the kind of information that is very difficult to obtain today and other means.
And at the time when XAI was releasing one of their earlier, or I should say, several models, we did not have those safeguards in place.