AI Safety Experts Guard Against Doomsday — But Not Your Job

An essay published on LessWrong — the online forum where much of today's "AI could end everything" movement got its start — makes a simple, uncomfortable point. The people most worried about artificial intelligence have built an impressive machine for preventing disaster: recruiting pipelines, funders, and researchers who deliberately took jobs in government and at AI companies to steer the future. What they have barely built is a plan for the opposite outcome: AI that works, spreads fast, and reshuffles the economy.

The author, writing under the name Spade, says that gap is backwards. If AI doesn't destroy us, it will very likely transform work — mass unemployment with "horrible knock-on effects," in their words. The stock market's current AI enthusiasm may already be a preview. Yet the small number of projects studying how jobs will change, and what governments or companies should do about it, is outnumbered "by a factor of 10x or 100x" compared with apocalypse-prevention work. The author flags that number as a rough guess, not a measurement.

There's a second worry: power. Cheap, capable AI gives single individuals abilities that once belonged to institutions — curing a disease, or building a bioweapon. Either way, the world becomes more volatile and harder for an ordinary person to plan around. A career, a mortgage, a retirement plan all assume the ground stays roughly where it is.

The author isn't arguing anyone should stop preventing catastrophe — obviously, survival comes first. The point is that surviving and then being miserable about it isn't much of a victory. The essay calls for more effort on the unglamorous questions: which jobs disappear first, how fast, and what actually helps people adapt without falling through the cracks. It's an opinion piece, not research, and the movement it critiques has published relatively little on this. But it names a blind spot worth watching.