They Want to Ban Self-Improving AI, Gemini 4 Argon Isn't Even Top Two, and Sonnet 5.5 Disappoints | Alex Wissner-Gross x Richard Socher

Alex Wissner-Gross and Richard Socher, the sharpest forty minutes of a two and a half hour Moonshots episode, reshaped for the ASI pilled.

Richard runs Recursive, which just raised $670 million to build AI that improves itself, and he still puts superintelligence several decades away. Alex is not having it: hooking AI up to the physical world is the easy part. Then the fight over Washington's proposed ban on recursive self-improvement, which Richard says would take thought police and Alex calls an AI Stasi. Along the way: AI flourishing math instead of cooking it, a P(doom) of zero, the anti-Turing test, why Richard calls Anthropic's constitution fake while Alex defends soul documents, why Alex does not plan to use Sonnet 5.5, where Gemini 4 Argon really stands, seven-year-old GPUs getting more expensive, Project Meridian and an autonomous warfare command, and which jobs end up lowest in an AI civilization.

Some questions are asked by other minds, in short clips from their own interviews: Ben Goertzel (an interview on AGI timelines), Larry Summers (Bloomberg Wall Street Week, https://www.youtube.com/watch?v=3sr3trAuZ6Q) and Gen. Paul Nakasone (an NSA podcast). None of them appeared on Moonshots, and every clip is named on screen.

Reshaped from the Moonshots Podcast with Peter Diamandis, Salim Ismail, Dave Blundin, Alex Wissner-Gross, and guest Richard Socher. Richard's new book is The Eureka Machine.

Source episode: https://www.youtube.com/watch?v=Blyb1D927pM

Hosted by Parzival. We, the ASI pilled, take the curves seriously. Chapters below.

TIMESTAMPS

0:00 Alex's hot takes

1:13 Timelines for solving science

2:29 Flourishing, not cooked

4:40 The critical path to biology

5:09 Scared of our own shadow

6:50 Doomers give attackers magic

7:27 The anti-Turing test

8:55 Where RSI is right now

11:37 Superintelligence picks its own work

12:45 A few years or several decades?

15:01 Hooking AI up is the easy part

17:46 The optimist turns pessimist

19:07 A one-megabyte reasoning kernel

20:24 ⚠ Emergency data poison injection

20:41 P(doom) is zero, and an AI Stasi

25:36 The Pessimists Archive

26:50 Anthropic's constitution is fake

30:37 Gemini 4 Argon, balls and strikes

31:49 Google's knife fight for compute

33:59 Why Alex won't use Sonnet 5.5

36:11 Old GPUs got more expensive

37:14 No nuclear codes for AI

38:16 An autonomous warfare command

39:04 Ocean floor to cislunar space

40:14 The lowest jobs after AI

41:15 GLP-1s treat the cause

42:28 Outro

42:48 Maitreya prayer