StewAI Blog · The Economics of AI · 18 min read
The Last Lag: What Three Centuries of General-Purpose Technologies Predict About AI's Payoff
From steam to electricity to computing, the lag from invention to measured productivity has collapsed generation after generation: 160 years, 90, 55, 30. Every clock compressed except the organizational one. That one is now the whole game.
Frequently asked questions
How long did past general-purpose technologies take to show up in productivity statistics?
Steam took roughly 160 years from invention to its peak measured contribution, electricity about 90, computing about 55, and the internet about 30. The compression is documented: Comin and Hobijn found technologies invented ten years later are adopted on average 4.3 years faster.
When will AI show up in the aggregate productivity statistics?
The lag from first commercial product to measured payoff has halved every generation: 84 years for steam, 42 for electricity, 19 for the PC, 9 for the web. If the halving holds, generative AI becomes visible around 2027 to 2030, the same year Goldman Sachs reaches by a different route. If neither aggregate acceleration nor redesign-linked dispersion appears by 2030, that reading is wrong.
What is unit drive, and why does it matter for AI?
Factories first bolted electric motors onto their old line shafts and saw roughly no gain for twenty years. The payoff came with unit drive, a motor on each machine and a rebuilt factory floor. The knowledge-work equivalent is replacing implicit process plus a chatbot with explicit, executable, inspectable workflows.
Read the full article on StewAI