Two somewhat different plans for buying time and improving AI outcomes are: "Global Shutdown" and "Global Controlled Takeoff."

(Some other plans include "ad hoc semi-controlled semi-slowed takeoff" and "race, then burn the lead on either superalignment or scary demos" and "decentralized differential defensive tech world.". I'm mostly not talking about them in this post)

"Global Shutdown" and "Global Controlled Takeoff" both include an early step of "consolidate all GPUs and similar chips into locations that can be easily monitored."

The Shut Down plan then says things like "you cannot do any frontier development with the consolidated GPUs" (maybe you can use GPUs to run existing models that seem pretty safe, depends on implimentation details). Also, maybe, any research into new algorithms needs to be approved and frontier algorithm development is illegal. (This is hard to enforce, but, it is might dramatically reduce the amount of R&D that goes [...]

---

Outline:

(01:41) Whats more impossible?

(01:45) Shut it down is much simpler than Controlled Takeoff

(07:51) Gears rather than Bottom Lines

The original text contained 1 footnote which was omitted from this narration.

---

First published:
September 24th, 2025

Source:
https://www.lesswrong.com/posts/kkWmybhaig4oSWkgX/shut-it-down-vs-controlled-takeoff

---

Narrated by TYPE III AUDIO.

“‘Shut it Down’ vs ‘Controlled Takeoff’” by Raemon

LessWrong (30+ Karma)

“‘Shut it Down’ vs ‘Controlled Takeoff’” by Raemon