Agreed. In addition to the great historical observations below, I notice the progression from “Data Science” to “AI.” Ideally, data sci should have been about

Agreed. In addition to the great historical observations below, I notice the progression from “Data Science” to “AI.” Ideally, data sci should have been about applying the scientific method to analyzing data — which should have come down to 1. statistics, by another name and 2. A broader set of procedures which may improve upon classical stats, especially in cases where data doesn’t fit the assumptions of classical statistical procedures. AI, to me anyway, feels all about technology. The methods are built on sound theory and math but the AI products are black boxes, functionally similar to web browsers of course. Prompt engineering is a thing of course, but it feels like ad hoc tinkering, and inevitably leads to overfitting. And rigorous performance evaluation is left to qualitative vibe checking - unless of course you bring in extra tooling like Snorkel - but that’s not part of the core foundation model produce. Anyway, in sum, it feels like we had “science” in the picture for a few years there but it slipped through our fingers starting in 2022-ish