When AI Makes Things Up (“Hallucinations”) The profoundly great power of “good enough knowledge” Every day, we’re inundated by information. Much of it arrives without much context. A lot of it is extremely complex. Some is produced in a real effort to inform, clarify, and make sense of the world. Some is designed to flatter or shame us into buying something, or fill us with doubt, or intentionally mislead us, or just distract us. Yet there are lots of settled truths (and mostly settled truths) out there, too, and I believe that having convenient access to this information has enormous value. Consider Wikipedia. Today, Wikipedia’s English version alone gets more than ten billion monthly page views from more than 850 million unique devices. Whatever amount of error it con- tains, I think it’s safe to say we’ve learned to live with it, and we now regularly depend on Wikipedia to help navigate and make sense of the world. How did this happen when, in Wikipedia’s early years, it was widely seen as its own kind of untrustworthy hallucina- tion machine? The site’s success is perhaps explained with a perspective that founder Jimmy Wales has often expressed about Wiki- pedia: “It’s good enough knowledge, depending on what your purpose is.” This resonates with a core principle I champion in my books and podcasts and that I almost always apply to my investing, political, and philanthropic decisions: good distribution is far more important to a product’s success than good service—or 159