Cory Doctorow says AI was built for capital, not workers

Cory Doctorow's new talk 'AI Was Never About Helping You', up on YouTube and threaded into a longer interview on Democracy Now this week, is the argument he has been sharpening for a while now, tightened into one sitting. The refrain he keeps coming back to is that the most important thing about a gadget is not what it does but who it does it for and what it does to the person on the other end of it, and he thinks the current AI build-out fails that test badly.

The central distinction is worth holding onto whether or not you buy the rest. When labor drives automation, Doctorow says, it tends toward making the product better, because the people closest to the work want better work. When capital drives automation, it tends toward making more of the product, often at the expense of quality, because the people paying for the automation want to amortise the asset. AI deployments, in his read, are mostly the second kind. The radiology example he keeps returning to is concrete: the pitch from AI vendors, as he frames it, is to fire nine-tenths of your radiologists to save roughly $2.7 million a year, and when the system misses something and a patient dies, designate the remaining human as what Dan Davies calls the 'accountability sink'.

He extends this to warehouse work, where he says the most heavily automated facilities produce three times the injuries of other warehouses, and to white-collar work more broadly through his 'reverse centaur' framing: instead of a human assisted by a machine, a human conscripted to keep up with one. The economic backdrop he puts underneath all of it is roughly $50 billion a year in AI revenue against $1.4 trillion a year in spending, with every new customer, on his telling, losing the vendors more money rather than less.

The honest caveat is that this is one critic's framing, delivered alongside book events for 'The Reverse Centaur's Guide to Life After AI', and the big numbers he cites are his to defend; neither the video page nor the interview I pulled audits them independently. The radiology savings figure in particular is a vendor pitch he is quoting back, not a measured outcome. What the reporting does not give you is a clean counterfactual for the deployments he is most worried about, or a concrete alternative beyond the principle.

For practitioners the useful thing to take from it is the labor-versus-capital lens itself. Tooling rolled out at the request of the people doing the work tends to age differently than tooling rolled out to shrink the headcount doing it, and the second pattern is the one Doctorow is betting will break first.