METR's David Rein argues that key parameters in current ODE models do not rule out a rapid AI intelligence explosion
Herbie Bradley argues the underlying literature remains largely unchallenged.
1. https://www.nber.org/papers/w35155 2. https://www.forethought.org/research/will-ai-r-and-d-automation-cause-a-software-intelligence-explosion 3. https://arxiv.org/abs/2309.11690 4. https://basilhalperin.com/papers/singularities_slides.pdf 5. https://arxiv.org/abs/2507.23181v2 6. https://arxiv.org/abs/2405.10494
david rein@idavidreinIt is an open scientific question whether an intelligence explosion is possible—the best estimates of the key parameters in the relevant ODEs do not rule it out. If you're confident "It Can't Happen", I'd be curious what you think of the literature in the next tweet:
@idavidrein I'm pretty confident good rebuttals can be written to a lot of the papers you link; IMO they mostly survive as the best estimates because few people have spent time on this—which itself is somewhat surprising
david rein@idavidreinIt is an open scientific question whether an intelligence explosion is possible—the best estimates of the key parameters in the relevant ODEs do not rule it out. If you're confident "It Can't Happen", I'd be curious what you think of the literature in the next tweet: