Historians are a lot better explainers than algorithmaticians. (Is there such a word? If not, I just made it up.) Let me argue my case.
Harvard’s Jill Lepore, one of the most fascinating and truly humorous historical writers, has superb explanatory insight on algorithmic prediction. That’s not a surprise. She approaches history from perspectives that have never crossed this historian’s mind--and does it with a lot of verve. She’s one of the few with a steady hand when the rest of us are lurching around with churning stomachs about artificial intelligence and moronic politicians.
With two talkative algorithmic geniuses in my own family, a son-in-law from Harvard and a recently minted Carnegie-Mellon data-scientist grandson, with whom I spend hours listening and making sense of their conversations, Lepore is an absolute delight at simplifying algorithmic insight.
She regularly seems to have the ability to condense an entire discipline, the whole shebang into a single readable paragraph. Here’s the first paragraph from her New Yorker article, entitled simply “Unforeseen.” Prophecy is a mug’s game. But then, lately, most of us are mugs. 2018 was a banner year for the art of prediction, which is not to say the science, because there really is no science of prediction. Predictive algorithms start out as historians: they study historical data to detect patterns. Then they become prophets: they devise mathematical formulas that explain the pattern, test the formulas against historical data withheld for the purpose, and use the formulas to make predictions about the future. That’s why Amazon, Google, Facebook, and everyone else are collecting your data to feed to their algorithms: they want to turn your past into your future.
Continue reading "Predictive Algorithms Actually Start Out as Historians" »