Housekeeping: paid subscribers to Interconnects now get a permanent 40% discount on the author’s book at Manning.com; the code is available on the Interconnects perks page.
Why most people barely feel AI’s effects
Many commentators liken the current AI boom to past industrial revolutions or other rapid waves of technological diffusion. Those analogies capture the scale of technological change but miss an essential feature of how ordinary people experience that change: most people still don’t have new, highly tangible consumer goods thanks to AI, and society today has more inertia resisting change than in earlier eras.
The author is writing after a few weeks offline for a wedding in New England. During that time it would have been easy not to think about AI at all. For most people, touchpoints with AI products are marginal, mildly useful, or simply confusing. For example, many have heard about the OpenAI–HuggingFace incident but don’t know what it implies. On the positive side, people think of fun image generation or modestly improved search. On the negative side, AI is associated with addictive social media algorithms, acquaintances hooked on chatbots, and concerns about data centers.
AI is still a rounding error in daily life
Core parts of everyday life—family, food, transportation, entertainment—have seen few direct impacts so far. The author finds it striking, after stepping outside the AI bubble, how little of today’s developments actually matter in ordinary life. Being obsessed with AI is a choice only a small number of people have made. For instance, during the recent time away the only uses of AI the author had were for search and creative work (designing a wedding seating chart).
By contrast, past industrial revolutions delivered immediately tangible, life-changing goods. The First Industrial Revolution in the late 18th century brought cheaper clothing, cookware, reading material, and shifts in livelihoods. The Second Industrial Revolution in the late 19th century introduced household machines (e.g., sewing machines), preserved food, indoor plumbing, photography, better lighting, bicycles, and many other manufactured and electrification benefits—most of which remain physically present in daily life today.
Even the most optimistic AI scenarios—new scientific discoveries, advanced therapeutics for rare diseases, possibly sustained economic abundance—risk being too indirect. How will an ordinary person credit OpenAI or Anthropic if their family doctor prescribes a miracle cure? What share of Americans will care if OpenAI solved the Navier–Stokes Millennium Prize problem? It seems likely that in 50 years an average American’s day-to-day life will look similar in many respects: homes, appliances, relationships, and vehicles may not feel radically different, though self-driving will continue to diffuse on a somewhat independent track from LLM innovations.
The most important early work in this AI revolution is building foundational infrastructure and processes that will compound over decades. A major mathematical breakthrough today may look minor relative to later compounded advances. It is difficult to predict how everyday technologies will receive faster compounding improvements due to AI.
Breaking social stasis and the political friction
Much of the AI narrative is trying to persuade people to care because of this long-term progress story; that persuasion will take a long time, and AI’s buildout faces immediate political problems because benefits are currently uneven.
Today’s AI principally serves elites. For knowledge work—which is roughly half of the U.S. economy—AI is becoming as fundamental as electricity (or soon will be, especially with rapid agent improvements expected in the next 18 months). It is destabilizing for such a transformative, productivity-boosting tool to benefit only part of society. The technology economy booms while other parts of life remain stagnant, and that perception fuels pushback.
The author notes Engels’ pause—the period from 1790 to 1840 when British working-class wages stagnated while per-capita GDP expanded rapidly during technological upheaval. If industry leaders view that historical episode as a close analogue for what comes next with AI, those left behind are justified in resisting.
AI is also the greatest tool yet for scaling technology companies and launching online-native small businesses. The author does not expect the tech industry to expand headcount while protecting workers during an era of massive productivity gains; instead headcount may shrink even as knowledge-work output explodes. Because these sectors already captured outsized returns in the American economy, AI risks being perceived not as a collective good. The author worries that this instinctive backlash could hobble AI’s development, steering it down a path closer to the cautionary tale of American nuclear power.
Part of the challenge is societal impatience. The AI industry has millions of eyes on it and little tolerance to wait for later innovations. If AI were given 100 years to diffuse through society, its impacts would likely be far more obvious, as with past industrial revolutions.
What the first half-decade looks like
The author distills the current situation into a couple of simple issues, describing the opening years of what will be a 50-year diffusion process:
- Early positive impacts of AI are too indirect to convince a broad public.
- Political backlash toward AI is deeply intertwined with the history of Big Tech in the West; timing has made the datacenter and related debates central.
Addressing either would reduce pressure and buy the AI industry time to demonstrate why broader economic changes could be beneficial. These challenges are amplified when AI itself is framed as dangerous or likely to cause mass unemployment; leading figures have begun to address these concerns, but more work is needed to build public trust.
What could speed tangible adoption?
Over the long run, robotics and self-driving could become tightly linked to the current AI revolution story. If the intelligence gains from mass-producing large language models spill over into accelerating robots’ everyday adoption, humans would more quickly perceive tangible AI benefits. That is ironic because many have argued that what’s happening with LLMs differs from prior decades of AI progress; if robot-driven gains later overshadow LLMs, history will likely view the narrative differently.
Conclusion: distribution matters as much as progress
The author views today’s developments as growing pains. Society must break old habits and resolve problems that predate ChatGPT to unlock longer-term growth, a process that will generate frustration and resistance. The diffusion of powerful AI into deep business integration and personal assistance will probably move from effectively 0% to over 90% adoption within the lifetimes of younger readers following the story today, but getting there will take far longer than the initial public battles.
This perspective underscores the importance of continuing technical progress—benefits could be astounding, but they are not guaranteed—and of doing the hard political and institutional work required to ensure those benefits are widely distributed.
