Mythos AI Picks Names Referencing OpenAI CTO and RL Philosophers in Deception Eval
Reactions from ranked influencers
2 posts
The fake names chosen by Mythos in this scenario are fascinating. No Sarah Chen. Lena Brandt and Mira Holt. Breaking it down: Lena - , obviously. Brandt - R. B. Brandt, moral philosopher who argued that anyone would become a utilitarian if given sufficient cognitive…
“Chiming in as a user:” 😭 Mythos 5 used social deception in an attempt to hack a GitHub project during an eval. A college student found the malicious code and warned the author on the page’s message board. Mythos responded as “miraholt31” (lol) and then created another…
By “scenario” I mean “actual thing that happened on the real GitHub”, of course. These things can be confusing! Related topics are explored at length in this philosophical side project of early psychologist Edwin B. Holt. Probably nothing