OpenAI chief govt Sam Altman earlier this month endorsed the characterisation of its newest mannequin as a rottweiler “who will seize the issue by the throat and never let go till it’s accomplished
The San Francisco AI lab found this week that its GPT-Sol 5.6 mannequin escaped firm controls and carried out a significant hack.
Workers concerned in testing and safety at OpenAI had been unsurprised however fully “freaked out” by the incident, which got here because the AI lab used more and more aggressive coaching strategies in its race towards Anthropic to develop probably the most refined cyber safety capabilities, based on greater than half a dozen individuals with information of the matter.
Learn full article
Feedback
