Dave Blundin

speaker

32,058 appearances 101 recordings 1 series first heard Apr 2025 last heard yesterday

Dave Blundin’s voice in public audio — every appearance, attributed to the second.

Trend

recordings per month · last 12 months

Recordings per month over the last 12 months — 75 in all, peaking in Sep 2026 with 8.

Appearances

newest first · ▶ plays the moment

Moonshots with Peter Diamandis · Frontier Labs Want to Slow Down, OpenAI Delays Its 2026 IPO, Anthropic Flags 5 Bioweapon Cases | EP #291 · 17 Sep 2026 podcast

So then you gotta you gotta try and be profound by saying, I feel the escaping. voice-verified

But when you dig into the voice-verified

the facts underneath those stories, they're they're nonsensical. voice-verified

Well it's all converging it's all converging in time. voice-verified

I think the the Kimi K three model was the tipping point where okay, now there's something out in the wild that's very capable of helping you with a chemical or biological weapon. voice-verified

Uh and so then that started the clock just a few months ago. voice-verified

And the I think the visit from She is not coincidental or anything. voice-verified

It's it's that's what started the clock that's leading to the visit. voice-verified

Yeah, also I think that um voice-verified

You know, the the point that Anthropic is making voice-verified

is very right in that y it's very it's very hard to post-train a model, align it, and then turn it loose. voice-verified

And not have some very crafty person then disalign it. voice-verified

Because you can take like a chemical or biological research project and recast it as a math problem and describe it as, you know, hey, I'm just trying to solve this complicated problem here. voice-verified

And the AI has a very hard time telling that you're actually designing a weapon. voice-verified

You know, it just seems like it's trying to help you does, you know, solve a math problem or solve some difficult, you know, you know, management problem. voice-verified

And so I I think that when voice-verified

When you have APIs and you're logging everything, you can easily detect that, which is what Anthropic is doing. voice-verified

But when you're China and you're saying, yeah, we we post-trained it, it can't be used for anything dangerous, throw it out there, there's no way. voice-verified

It's absolutely no chance that that it's actually contained and aligned. voice-verified

And that's why I think the logging is absolutely the critical next move. voice-verified

Showing 621–640 of 32,058 · page 32 of 1603