hn.today

On The Brutality Of Amdahl's Law

bosoncutter.substack.com22 points5 comments
Screenshot of On The Brutality Of Amdahl's Law

The central argument is that rapidly improving AI will not automatically produce equally rapid economic change. Amdahl’s law shows why: speeding up one stage of a process yields limited gains when other stages remain slow. In a hierarchy of bottlenecks, even making the current constraint infinitely fast produces only a constant-factor improvement; repeated breakthroughs can compound into exponential progress, but only if intelligence repeatedly becomes the next bottleneck. The key question is therefore what limits a task after AI makes analysis and decision-making cheap, fast, and highly capable.

Many constraints lie outside intelligence. In a factory, AI might identify the source of a defect and choose a fix, but testing materials, ordering parts, installing equipment, and verifying production still take time. Shipping, manufacturing capacity, regulation, and political will likewise resist instant acceleration, while scaling up requires scarce components and new infrastructure. Real-world uncertainty also matters: in fields such as biology, experiments can be slow, variable, and costly, so crucial information may take a long time to obtain even with excellent reasoning. The argument is that these bottlenecks help explain why capable agents have not yet caused sweeping economic disruption. Past general-purpose technologies transformed work and daily life, yet US GDP per person grew at roughly 2% annually over the long run; AI’s broader impact must be judged against such limits, not intelligence alone.

Read on bosoncutter.substack.com5 comments on Hacker News

Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.

More in AI

The daily digest

Today's best Hacker News stories, summarized and screenshotted, one email a day.