Artificial Intelligence Resources

Cory Doctorow and Tim O’Reilly discuss “enshittification” and the future of AI. Cory’s notion of enshittification describes how online platforms win over millions of users by providing great service and delighting their users, but once they’ve established their position, they begin to prey on the new market they created. First, they come for their suppliers and advertisers, but eventually they make services worse and worse for their users in order to extract more profit. O'Reilly has developed very similar ideas using the theory of economic rents. He notes that search engines, ecommerce sites, and social media are all “algorithmic attention allocation engines” and has documented how they can extract rents by tuning their algorithms and designs for their profit rather than user or supplier benefit. Both Cory and Tim worry that this same pattern will repeat itself for AI. Right now AI companies are in the virtuous phase of surprising and delighting their users. Will this continue once they’ve fully established themselves? We need to think about how enshittification will play out in new kinds of products we’ll be building, realizing that AI is inevitably a tool that will be incorporated into everything: our cars, our calendars, our news, our VR goggles, and many products we haven’t envisioned yet. Is it possible to use open source AI to build products that aren’t manipulated by the established monopolies? Can AI be used to “move value to the edges of the network,” as Cory writes? Can transparency and disclosure about how AI is used—not just that it is used—alert users, investors, and regulators to enshittification before it’s too late? And what kinds of transparency would inform users and encourage them to respond effectively?