Economist Alex Tabarrok makes a point so simple it's almost annoying. A 40% unemployment rate and a 3-day work week are the same thing. Same total hours worked, different distribution. So whether AI creates a catastrophe or a paradise depends mostly on policy choices, not technological destiny.
Between 1870 and today, US work hours dropped roughly 40%, from nearly 3,000 hours per year to about 1,800. Unemployment didn't spike. People just worked less over their lifetimes. In 1870, about 30% of your life went to work. Now it's closer to 10%. Childhood got longer, retirement got longer, people lived more. We've already absorbed a massive shift toward leisure once.
Tabarrok argues that AI is a general-purpose technology, one that will boost productivity across skill levels rather than rendering whole populations unemployable. And we have real policy levers: an "AI dividend" funded by taxes on automation gains, more mandated holidays, shorter standard work weeks. The Alaska Permanent Fund already shows the dividend model works, paying residents annually from oil revenues. Bill Gates proposed a version of this with robot taxes back in 2017.
The skepticism is fair. Hacker News commenters pointed out that economists got the web's impact wrong, so why trust their AI forecasts? Fair point. But Tabarrok's core observation holds regardless of prediction accuracy. The real question is whether we'll distribute AI's disruption like a layoff or like a vacation.