Full Transcript: What Anthropic Doesn't Want the Pentagon to Have
Real, documented, sourced AI capabilities that most people have no idea exist yet, as explained in today's video report.
This transcript has been lightly edited for concision and clarity. See the video edition here:
The Pentagon is trying to bully Anthropic into giving them a guardrail-free version of their technology to do whatever they want with. As Defense Secretary Pete Hegseth has put it: “We want AI that fights and wins wars. We don’t want woke AI.” So I decided to look into the very capabilities that I think Anthropic is most concerned its AI might be used for.
A few months ago I had Aza Raskin, co-founder of the Center for Humane Technology and one of the main guys behind The Social Dilemma, on the show. He and I agreed that there’s been so much dark talk about what AI could conceivably do to us all, and because I spent so much of The Loop — my book about AI — in that mode, I know now that it’s not for everyone. There’s a fatigue that grabs people when you try to scare them too often and too deeply.
But as Aza put it to me at the time, there’s some stuff going on right now that really can fuck up your sleep. So as we stand on the verge of Anthropic’s deadline from the Pentagon to give them their guardrail-free capabilities — what the Pentagon is calling “all lawful uses,” but that could include all sorts of things we don’t know about — I decided to remind people what this technology can actually do.
