AI Can Now Do Harvard-Level Work. What Happens to Jobs Next?

In this clip, Jason Furman revisits one of the oldest fears in economics: that new technology destroys jobs. He explains why, for roughly 150 years, that fear has almost always been wrong. Drawing on farming, accounting, and radiology, he shows how automation typically reshapes jobs rather than eliminating them. When spreadsheet software made accounting far more efficient, firms didn't hire fewer accountants. They hired more, because real demand for accounting work was far larger than expected. Furman breaks from that historical confidence here, though. He describes going from one full-time research assistant to zero as AI now handles that work, and argues today's models have reached something close to "artificial college intelligence," able to complete any Harvard-level assignment, often better than students can. Still, jobs are bundles of many tasks, not single tasks, so eliminating homework-style work isn't the same as eliminating a job. The pace and breadth of this shift is different enough to warrant real caution. Jason Furman's presentations highlight: • Why 150 years of automation history shows technology rarely eliminates whole professions • How accounting demand grew, not shrank, after spreadsheet software automated the work • Why radiologists are more common today despite AI's ability to read scans • What separates a "task" AI can do from a full "job" made of many tasks • Why this wave of AI feels different from past waves of automation panic Key takeaway: Automation has historically reshaped jobs rather than eliminated them, but the speed and breadth of current AI capability is enough to question whether that pattern holds.