How worried should we be about AI?

Jerusalem Demsas and Matt Yglesias debate whether artificial intelligence poses an existential risk to human civilization, or if that idea is just a dangerous distraction from real, near-term risks. Matt's case for doom is that we'll cede more power and authority to AI agents as they recursively self-improve, leaving humans more vulnerable in the end. Jerusalem pushes back on this and argues that obsessing over extinction scenarios makes it harder to regulate biological weapons, cybersecurity vulnerabilities, and the economic disruptions already underway.

Plus a peer review of a new paper on why meetings can actually be good for you and your company.

(00:00) Is advanced AI going to kill us all?
(04:09) Jerusalem pushes back on AI doom
(07:26) The Claude Code auto-mode problem
(11:02) Bioweapons and AI's real near-term risks
(15:41) Why extinction risk still matters politically
(19:21) Why Matt is scared of a flawless AI
(33:13) Reasons AI won't take over
(52:08) The case for an AI pause
(58:54) Peer review: why meetings are actually good for you

Support the Show: Get a full transcript, show notes, and an ad-free version: TheArgumentMag.com

New episodes every Thursday — subscribe so you never miss a debate!

The Argument is produced by Justin Zuckerman, fact-checked by Eli Richman, with music by Breakmaster Cylinder and art by Ben Tousley.