The Future of Work in the Age of AI

Blue-collar work has long been reshaped by machines: farmers and factory workers now produce far more with a fraction of the labor force they once needed. The question raised here is whether the same forces are coming for white-collar work, the tasks performed by lawyers, doctors, teachers, accountants, architects and consultants, and why so many people resist that idea even as the evidence mounts.

Examples drawn from well before the arrival of ChatGPT illustrate how far this had already gone: an online dispute resolution platform handling far more cases each year than the entire US and UK legal systems combined, a diagnostic system matching dermatologists at spotting cancerous freckles, a concert hall designed by an algorithm, and a Vatican-approved app for preparing for confession. These cases share a pattern: complex, judgment-heavy work being performed by systems that understand nothing about the field they operate in.

A history of artificial intelligence explains why this is possible. Early AI, built in the 1980s, worked by capturing a human expert's reasoning in a giant set of rules, an approach that stalled when experts couldn't explain their own instincts. Garry Kasparov's defeat by Deep Blue in 1997 marked the shift: the machine won not by thinking like a chess grandmaster but through brute-force calculation and processing power nobody had anticipated growing so fast.

That history exposes what's called the artificial intelligence fallacy, the mistaken belief that a machine can only match human performance by copying human methods. Applied to professional work, this reframes the usual defense that certain tasks require judgment. The real question isn't whether a machine can exercise judgment, but whether it can handle uncertainty better than a person can, and in many domains the data says yes.

• Historical data from agriculture and manufacturing showing output rising even as the workforce needed to produce it shrank
• Case studies of professional tasks already automated in law, medicine, and design before generative AI existed
• An explanation of why 1980s-era AI failed and what changed by the time Deep Blue beat Kasparov
• A reframing of "judgment" as a response to uncertainty, and why that changes the automation debate
• A closing argument for education and training as the necessary response to these shifts