Why most people still don't feel AI's impact

Nathan Lambert makes an uncomfortable point: AI is advancing fast on benchmarks and in research, yet for most people it remains, in his words, "a rounding error in everyday life." The touchpoints an ordinary person has with AI are still fringe, marginally useful, or simply confusing. His argument is that the benefits are real but indirect, so the public has little reason to attribute better healthcare or cheaper software to AI, and that gap breeds political backlash before it breeds gratitude.

His historical comparison is the sharpest part. Past industrial revolutions delivered things people could see and hold: cheaper clothing, household machines, preserved food, indoor plumbing. AI's early payoffs are scientific discoveries and rare-disease therapies, which are diffuse and hard to trace back to a chatbot. He reaches for the "Engels' pause" of roughly 1790 to 1840, when British output grew quickly but working-class wages stagnated for decades. AI today mostly serves an elite of knowledge workers and the companies that employ them, which is exactly the pattern that produces resentment rather than broad support. Lambert's timeline is deliberately long, on the order of fifty years for real diffusion, with robotics and self-driving cars as the technologies that could finally make the benefits physical and obvious.

Why it matters

If you work in AI, this reframes the "why doesn't the public get it" complaint: the problem is not messaging but the shape of the benefits, which stay concentrated. If you plan products or policy, the useful bet is on things people can touch, and on preparing for the stagnation and backlash that a long diffusion period tends to bring.