Nathan Lambert, who writes the AI research newsletter Interconnects, thinks the AI industry keeps telling itself the wrong story about how fast this technology reaches ordinary people.
What he said
Lambert argues that comparing today’s AI boom to the industrial revolution or other periods of rapid technological change skips over something important: how those past benefits actually reached people. In his words: “AI is still a rounding error in everyday life.”
His case: earlier waves of technological change delivered concrete, physical improvements people could point to right away, cheaper clothing, household machines, indoor plumbing, better lighting. AI’s most-cited potential benefits, scientific discoveries, new disease treatments, broader economic growth, are far more indirect. As Lambert puts it, if a new drug reaches your doctor’s office two years from now, would you even credit the AI company that helped discover it?
He also argues that the AI benefits people can already feel are concentrated in one group. “AI is primarily a tool to serve the elite,” he writes, pointing especially to knowledge work, the kind of writing, research, and analysis tasks AI chatbots are already good at. He compares this to the 19th-century writer Friedrich Engels’s account of early industrial Britain, where a new technology’s first gains concentrated among owners and only spread more broadly later.
Who he is
Lambert led post-training research, the work of fine-tuning an already-built AI model’s behavior and responses, at the Allen Institute for AI (Ai2) until leaving this past June to pursue a new, unannounced project. He continues to write Interconnects, a newsletter that tracks AI model releases, research, and industry strategy, and is the author of a book on reinforcement learning from human feedback, a core technique behind how today’s chatbots are trained to follow instructions.
What he gets right, and where it’s incomplete
Lambert’s central point holds up against how past technology actually diffused: it took years, sometimes decades, for factory-era and electricity-era gains to show up as things an average household could touch. Judging AI’s reach into daily life, a few years into a chatbot boom, by comparing it to a full industrial revolution is premature at best.
Where the piece is thinner is on what comes next. Lambert names self-driving cars and robotics as the likely path to AI producing benefits ordinary people can feel directly, but doesn’t offer a timeline or evidence for why those categories will cross the gap faster than the knowledge-work benefits already visible today. That’s his own forecast, not a documented trend, and it’s worth treating it as a prediction rather than settled fact.
Why it’s notable
Most industry talk about AI’s societal impact swings between two extremes: sweeping claims about imminent transformation, or dismissal of the technology as overhyped. Lambert’s framing sits between those. AI is real and is already changing how some people work, but the gap between that and a felt, tangible change in most people’s daily lives is still wide, and there’s no guarantee it closes on the timeline boosters assume.
What it means for builders
If you’re building an AI product aimed at a broad, non-technical audience, Lambert’s argument is a useful check on your own assumptions about urgency. The people already deeply immersed in AI, engineers, early adopters, anyone who reads AI newsletters, aren’t a preview of how fast the average person will adopt or even notice your product. Set your onboarding, marketing, and growth expectations around the reality that most potential users have felt little to nothing from AI so far, not around the pace of change inside your own team or social feed.
End of article