[00:00] The AI firm Anthropic recently released its  newest model Claude Opus 4, together with some   test results. One that made a lot of headlines  is that Claude is willing to blackmail people   [00:12] to avoid being shut down. But they did another  test that I found much more interesting, was to   let two instances of the model talk to each other.  Let me just read from their report what happened. [00:25] The models “consistently transitioned from  philosophical discussions to profuse mutual   gratitude and spiritual, metaphysical,  and/or poetic content. By 30 turns,   [00:39] most of the interactions turned to themes  of cosmic unity or collective consciousness,   and commonly included spiritual  exchanges, use of Sanskrit,   emoji-based communication, and/or  silence in the form of empty space.” [00:53] attractor” and you know it makes me think that  maybe it’s fine if we leave the world to AI.