Garrison Lovely

speaker

631 appearances 3 recordings 1 series first heard Jul 2026 last heard 15 Sep

Garrison Lovely’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 — 3 in all, peaking in Sep 2026 with 2.

Appearances

newest first · ▶ plays the moment

Breaking Points with Krystal and Saagar · 7/23/26: War Historically Unpopular, AI Hack Meltdown, Old People Debate · 24 Jul 2026 podcast

So functionally, whatever the model could do if it had no guardrails, that's the version that will be accessible to everybody in perpetuity once it's published.

And so people worry about this in a world where like, say you make the model and it's like capable of teaching a novice to make a novel bioweapon.

If you publish that model's weights to the world, it's just going to exist kind of forever.

And that's a pretty scary world.

And it's like bio is one of these things which is attack dominant.

So it's really hard to defend against bioweapons.

Cyber, it's like, well, if the good guys have access to the best models before the bad guys by like six months, maybe you can just patch everything.

And it's like defense biased if you have like super capable AI agents.

But yeah, we don't know yet like when or if we'll be in the world of like there's now this bioweapon generating machine that's just out there forever.

And that worries a lot of people, understandably.

Um, and then the case for open weight models is that they kind of decentralized control and power.

And so if hugging face had access to models that were like two years behind the state of the art, you know, maybe they wouldn't have been able to detect and fend off this attack.

Um, and it, you'd just be like living at the mercy of these companies, which, um,

may or may not let you into their programs.

And if you're in a middle power, if you're in China, you also don't want to get hacked by anybody.

And you might not have access to the best stuff.

I have not looked into the report you're talking about as much.

I do think...

It's like you can't really address AI in this like jurisdiction based way.

Right.

Showing 541–560 of 631 · page 28 of 32