The first protest I ever went to was with PauseAI US, thanks to Holly Elmore. Liron Shapira was there with a megaphone, leading the chants and handling the media, and I learned something that day: a movement needs people who understand the arguments, but it also needs someone who can say them out loud, clearly, to a crowd and a camera. Liron does both, which is more impressive once you know he’s an introvert who finds all of it draining.
He hosts Doom Debates, where he’s debated AI risk with Robin Hanson, Vitalik Buterin, Gary Marcus, Max Tegmark, and Dean Ball. He’s been reading Eliezer Yudkowsky since 2007 (he describes his show as spoon-feeding Eliezer from twenty years ago to people who are still catching up), and of everyone I’ve talked to, including the leaders in the field, he’s the person who thinks most like me about all of this.
His argument is the one in the title. The people running the frontier labs believe they’re building something godlike, and Liron thinks they’re right. The disagreement is about what happens next: they assume it will keep listening to us, and he wants to know why they’re so confident. Once a technology outdoes the brain the way an airplane outdoes a bird, nothing we care about is guaranteed to survive contact with it. His p(doom) is around 50 percent, which he calls the honest middle: not 1 percent and not 99 percent, since either one bets the species at absurd odds.
“I don’t wake up in the morning feeling like I need hope. I feel fine. I just wake up feeling like I should keep pointing out that, logically, we don’t seem fine.”
We talked about why the labs build it anyway, recursive self-improvement and the 2031 consensus timeline, why people as brilliant as Roger Penrose and David Deutsch can miss what he considers a basic argument, what it’s like to believe this while everyone around you expects another decade of business as usual, whether the rationalist community is a cult, and what he’d do if he were king of the world (pause frontier training, which he hates saying, because he leans libertarian and is having a great time using these models).
We then talked about why he has kids despite all the risk.
My co-interviewer Lei Huang took the second half, got Liron to build the extinction argument from scratch, and then pushed on the harder personal questions: whether he’s monetizing fear (he takes zero dollars from Doom Debates and pays the bills with a relationship-coaching company he’s run for ten years), why so many smart people disagree about something this important, and whether he feels like a chosen one.
Watch: Silicon Valley Thinks It’s Building God. He Agrees. | Liron Shapira
Chapters
0:00 Cold open
0:51 Who is Liron Shapira? Doom Debates and 19 years of reading Yudkowsky
4:25 Silicon Valley is building God — and why they build it anyway
6:59 What’s your p(doom)? Gradual disempowerment vs. “cancer AI”
9:18 Timelines (2031) and recursive self-improvement
10:47 Skeptics, the Erdős problem, and why geniuses miss the argument
15:57 Feeling disconnected from normal people; the spectrum from job loss to extinction
18:12 If he were king: pause frontier training, stop buttons, and a libertarian’s discomfort
21:35 Who Liron actually listens to
22:45 Growing up in Silicon Valley, moving to upstate New York, being on the spectrum
24:38 What his wife thinks; the “why not take out a huge loan?” gotcha
27:43 Why isn’t his p(doom) as high as Eliezer’s? Von Neumann and nuking Russia
29:54 Should you have kids?
30:31 S-risk: outcomes worse than extinction
32:17 What gives him hope?
33:31 Ranking the lab leaders; is alignment solvable with enough retries?
35:23 Is violence on the table?
36:52 Is the rationalist community a cult?
37:52 Rapid fire: The Truman Show, the smartest person alive, love, Satoshi, cryonics, the simulation
42:07 The extinction argument from scratch: airplanes vs. birds
45:27 The second crux: what happens when everyone has a superintelligence?
48:46 Why do smart people disagree? Epistemic vertigo, incentives, biases
52:42 What Doom Debates is for: cross-examination as a missing institution
55:09 The day job: Relationship Hero
58:19 “Monetizing doom”? Why Liron takes $0 from the show
59:56 Do you feel like the chosen one?
1:02:57 Successionism: is humanity just a bootloader?
1:04:18 How he picks guests; coming late to the party; what’s next
1:08:00 Outro
Where to find Liron
Doom Debates on YouTube: youtube.com/@DoomDebates
Doom Debates on Substack: doomdebates.com
Liron on X: x.com/liron
I’ll end the way the video does. Are you convinced? If not, tell me which argument you’d need to see addressed. If so, tell me what you think we should do about it. And if you know someone who hasn’t heard the case yet, this is the episode to send them.
If you liked this interview, definitely check out this one:
Thinks it's building something "God-Like"???
Want to align AI?
Reverse that!