I’ve been broadly bullish about AI, and regular readers here know that. The tools are useful already, and designers who learn to direct them will have a real advantage.

That is why I want to make room for Matthew Butterick’s more pessimistic argument. Butterick is a typographer, writer, programmer, and lawyer who has been directly involved in legal challenges to generative-AI training practices. His argument belongs in the AI conversation precisely because it does not require a rogue model or a Skynet story:

Among AI risks, we should take more seri­ously the poten­tial conse­quences of AI working as intended. AI is a capi­talist instru­ment. Its prin­cipal func­tion is to concen­trate capital. Its intended mech­a­nism is large-scale labor replace­ment. But it is also inher­ently polit­ical tech­nology. As AI makes it harder for workers to capture value from their labor, they will increas­ingly have to rely on goodies from Big AI, priva­tizing what were once func­tions of govern­ment. If Big AI subsumes the func­tions of workers and govern­ment, both will tend to realign polit­i­cally around Big AI’s inter­ests. What­ever term describes this system, it is not liberal democ­racy as US citi­zens have tradi­tion­ally under­stood it. AI-centered capi­talism risks an extinc­tion of demo­c­ratic possi­bility. It will be America. But it will no longer be Amer­ican.