The Industrial Revolution Is the Wrong Precedent for 20% AI Growth

Taking a precedent seriously means testing what it actually implies, not borrowing its romance. The Industrial Revolution is the headline historical case for explosive growth, so it gets cited whenever someone argues AI can push annual growth above 20%. Apply the same standard-deviation math that makes the IR look extraordinary, though, and the precedent lands somewhere far more modest. Matt Clancy's working paper using the Maddison Project database shows that a 10x acceleration over the long-run pre-IR rate was already a common good year in 1252 to 1652 England. Re-run that arithmetic on the modern frontier and you get roughly 2.8% annual growth, not 20%. Believing the 20% case is plausible therefore means conceding that today's economy would have to be qualitatively different from any year on the modern record, an order of magnitude beyond what the IR's own logic licenses. The "IR precedent" is doing less work than its proponents claim. It is a benchmark, not a license, and the benchmark says modest acceleration, not regime change.