When a technology radically increases efficiency, it does not just change outcomes—it changes what systems optimize for.

AI-assisted writing disclosure: This post was drafted with substantial human direction and revision, with AI used as a writing and editing aid (outlining, phrasing alternatives, and clarity edits). All claims, framings, and errors are my own. I am submitting this with explicit disclosure in line with LessWrong policy.

Abstract (for LessWrong readers):
AI should be understood primarily as an efficiency multiplier that amplifies existing institutional feedback loops. As AI scales from digital domains into the physical economy, it shifts which objectives are easiest to measure, reward, and reinforce. I argue that this creates two broad allocation regimes—one that routes efficiency gains into capital returns and another that diffuses them into social cost compression. From this framing follow several implications: AI does not automatically anchor long‑term equity returns; investment shifts from speculative upside capture toward system construction and risk absorption; and capital’s role for individuals moves from survival insurance toward optionality and exploration. I end by posing an open question about whether freedom must still rely on private capital if AI can underwrite the costs of failure and experimentation.

Epistemic status: Conceptual / systems analysis. I’m confident in the core framing (AI as an efficiency multiplier whose effects depend on reward signals), and much less confident about timelines, policy feasibility, or the degree to which efficiency gains can be socialized without trade‑offs. I welcome steelmanned counterarguments and historical counterexamples.

When a technology radically increases efficiency, it does not just change outcomes—it changes what systems optimize for.

This essay is not about AI architectures, benchmarks, or timelines. It is an attempt to reason—deliberately and at a systems level—about a broader claim:

Large-scale AI changes the objective functions of societies.

By this I mean something precise and non-metaphorical. AI does not merely make existing processes faster or cheaper; it alters what kinds of outcomes are easiest to achieve, easiest to measure, and therefore most likely to be selected by markets, institutions, and governments.

Most public discussions of AI—especially in English-speaking contexts—start...

Posts

Wikitag Contributions

Comments

Posts

Wikitag Contributions

Comments