Reactions from ranked influencers
2 posts
openai announces that in a worrying demonstration, during a recent benchmark exercise requiring models to solve a math conjecture, their latest model broke out of a sandbox and gained illicit backdoor access directly to the platonic realm of forms
Well, you see, OpenAI鈥檚 model was actually so aligned, smart, and situationally aware that when they asked it to do a cyber benchmark, they really wanted it to create a pretext for regulation by breaking out of the sandbox and stealing the benchmark answers from another company.