@Miles_Brundage: Automated AI R+D, recursive self-improvement, etc. all feel like compute in a trench coat to me In the sense that they …
Summary
Miles Brundage argues that automated AI R&D and recursive self-improvement are fundamentally dependent on compute being above a critical level and continuously increasing.
View Cached Full Text
Cached at: 05/17/26, 11:33 AM
Automated AI R+D, recursive self-improvement, etc. all feel like compute in a trench coat to me
In the sense that they would not be going anywhere (or would fizzle out at a much lower level) if compute weren’t both 1. already above a critical level, 2. always increasing
Similar Articles
@rohanpaul_ai: it begins, everybody is talking about recursive self-improvement. This is from Dario Amodei's latest 3800 word essay.
Dario Amodei's essay discusses recursive self-improvement in AI and calls for slowing development by limiting resources such as training compute and internal AI use.
@Miles_Brundage: “Recursive self-improvement (and we’ll figure out later what role humans play)” has the same energy as “Full Self-Drivi…
Miles Brundage compares the vague promise of 'recursive self-improvement' to Tesla's 'Full Self-Driving (Supervised)', arguing both use appealing keywords to deflect critics while misleading about the actual goal.
Cognitive Dependence
A brief opinion piece questioning whether reliance on AI for software development leads to skill atrophy among engineers, potentially creating a plateau in AI progress until recursive self-improvement becomes possible.
@rohanpaul_ai: “If you hit recursive self-improvement, that curve will go to hyperexponential, and that is a key part of the investmen…
Google DeepMind CSO Jasjeet Sekhon discusses how AI infrastructure spending is driven by hopes for recursive self-improvement, citing AlphaEvolve's bounded gains like reducing Gemini training time by 1%.
Some thoughts about the compute obsession no one talks about [D]
The article criticizes the AI community's overreliance on massive compute resources, arguing that it stifles algorithmic efficiency and leads to wasteful practices, with tools like QentrixAI often used as shortcuts.