Generative AI vs. the PC and the Internet
Three technologies on one clock: the share of people using each, counted from the year its first mass-market product appeared. The IBM PC took about twelve years to be used at work by 45% of employed adults. Generative AI got there in three and a half, on the same survey question. The dashed internet line is the share of all adults online, a broader measure shown for scale. The generative AI series is young and self-reported, and a tool that is free to try spreads faster than one that had to be bought and installed. The comparison is about pace, not about what the tools have changed yet.
What does it show?
Pace is the story, not level: the PC's curve was the fastest anyone had seen, and generative AI is running about three times faster on the same question.
Methodology
Three technologies on one clock: years since the first mass-market product — the IBM PC (August 1981), the commercial internet (April 1995, when NSFNET was decommissioned) and ChatGPT (November 2022) — the framing of Bick, Blandin and Deming's Generative AI Adoption Tracker (St. Louis Fed, 'The Rapid Adoption of Generative AI'). The two at-work lines answer the same survey question forty years apart: the Real-Time Population Survey reused the CPS Computer and Internet Use supplement's wording ('Do you directly use a computer for your job?') for generative AI, so the comparison is between employed adults who use the technology for work. PC points are the published Census Bureau figures for 1984, 1989, 1993 and 1997 (employed individuals 18+) and BLS figures for 2001 and 2003 (employed persons 16+), the years the supplement asked; the line is drawn between them. The generative AI line is the RPS quarterly adoption rate for work, from FRED, placed at each wave's survey month. The dashed internet line is a different population — the ITU's share of all individuals using the internet, via the World Bank — shown for scale, as the tracker itself does; it overstates workplace use in the early years. Adoption here means having used the technology, not using it daily: see the intensity chart for that. The generative AI series is young and self-reported, and a technology that is free to try will always be adopted faster than one that had to be bought and installed.