Artificial intelligence models learn from the finished products of human work, not from the process that produced them. Paintings in a gallery hide the practice and failed attempts, recordings leave out bad takes and jam sessions, books omit drafts, and scientific papers leave out years of dead ends. What AI sees is the trace of thinking, and what it produces are traces shaped like those products, without the experience behind them. The concern is that these systems can imitate expert output convincingly without having gone through the trial and error that builds judgment. If people lean on them and skip that struggle themselves, human expertise could be thinned too, since expertise comes from doing the messy process, not from reading the result.

In 1909 D. W. Griffith made a 14-minute film about a speculator who corners the wheat market and doubles the price of bread. A farmer sows his field while a 'Wheat King' gambles in the Chicago wheat pit; bakers raise the loaf from five cents to ten and the poor line up for bread. Griffith cut back and forth between the farm, the feast and the breadline, a technique called parallel editing, so audiences saw both ends of the economy at once. The film drew on Frank Norris's stories and on a real scheme: months earlier Chicago financier James Patten had cornered wheat, sent bread prices up and made about $5 million, roughly $185 million today.