Nathan Livni

AI Adoption and Milo's Cotton

The rush to put AI into everything reminds me of Milo Minderbinder cornering the Egyptian cotton market in Catch-22. Milo buys all the cotton because it seems like a brilliant deal. Once he owns it, he discovers that nobody wants it. He cannot admit that he made a bad investment, so he tries to make everyone eat the cotton.

There is something similar in the way the United States deals with corn. We produce an enormous amount of it, helped by a whole system of subsidies and incentives. Then we find somewhere to put it. Corn becomes sweetener, animal feed, fuel, starch, and filler. Its presence in everything does not prove that people demanded more corn. It proves that we had a lot of corn and powerful reasons to use it.

AI adoption may follow the same pattern. Companies have invested enormous amounts of money in chips, data centers, models, and startups. That investment now needs to become justified consumption. AI appears in search, email, customer service, education, toys, appliances, and software that worked perfectly well without it.

This looks like supply and demand, but the demand is not independent. The people who built the supply can manufacture demand through bundling, defaults, workplace mandates, advertising, and removing the non-AI option.

The loop:

  1. Institutions invest heavily in AI.
  2. They need widespread use to justify the investment.
  3. They put AI into products and require people to use it.
  4. The resulting usage becomes evidence that people wanted AI.

Everyone inside the system may recognize the absurdity while still having a reason to participate. Executives need an AI strategy because investors expect one. Workers use AI because executives bought it. Customers encounter AI because companies need usage numbers. The usage numbers then support another round of investment.

Everybody gets a share. And everybody has to eat the cotton.

Questions to pursue