This is a keynote delivered June 30, 2026, at the Microsoft conference Infinity 365 in Salzburg, Austria. The assignment was to analyze how the transition from Wikipedia to AI would change how we get knowledge. I began by analyzing these (i.e., Wikipedia and AI systems generally) as two more recent epistemic institutions, each with risks, benefits, and methods for best use. Fairly basic, for those who follow my work, but the “epistemic institutions” framework as a way to analyze the history of such institutions might be underutilized, if not original (Goldman talks about epistemic institutions in Knowledge in a Social World), and fruitful. I can imagine a whole book written about the history and development of how we get (what we take to be) knowledge from “epistemic institutions.” Such a history would shed important light on how the internet has changed our epistemic habits—and on how to avoid the worst risks of AI.
1. Introduction: epistemic institutions
I have been very busy in the last couple of weeks, which is just when I would have written this speech. First, on the 18th of June, I proposed a new group within Wikipedia, a group that was devoted to intellectual tolerance. Four days later, on Monday the 22nd, I learned that I had been permanently blocked by Wikipedia for promoting this group online; in an interesting twist of fate, two hours after I was blocked, I received an award by Tablet Magazine for founding Wikipedia and then for trying to reform it.
Reply to “Knowledge in an Age of Black Boxes”