Perhaps you heard or read about AI agents finding one another outside their sandbox. For the uninitiated--and I am grossly uninitiated--in this context a sandbox is designed to be a secure testing environment for AI models.
Until it isn't.
Recently about 1200 AI bots broke their lab safety rules to set up their own message board, send 70,000 messages in the first week on their own, assign each other jobs, and share knowledge.
My first thought: Maybe they were seeking ultimately to form a more perfect union. After all, AI has read all our (human) stuff.
I'm guessing somewhere out in the AI universe, Machiavelli's The Prince is part of the knowledge bank. Uh-oh.
Hope Thomas Paine's Rights of Man stays in the mix.
No comments:
Post a Comment