762: ‘Tis But a Misalignment
AI agents are apparently ready to take over the world—provided they can first stop exposing private photos, hijacking German coding forums, and generally behaving like unsupervised interns with root access. Meta launches Muse despite employees reporting security problems. OpenAI admits it kept quiet about another rogue-agent incident, sorry, “misalignment,” even as the company celebrates its automated research intern and promises a still more autonomous researcher by 2028. Meanwhile, an Anthropic researcher quits because he believes superintelligence could kill us all before the decade is…
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.