Back Then a Model Only Stole My Movie Ratings for Collaborative Filtering

neal_lathia · x · 2026-09-12

neallathia offers a wry take on the model-training-data debate: in the old days, when an algorithm "stole" his data, it just learned his film ratings for collaborative filtering recommendations. Today models take the same data and use it to solve math problems. The quip highlights the stark contrast between the recommender-system era and the LLM era in how personal data gets used — and implicitly suggests the current data-scraping controversy is an old story at a very different scale.

Original post →

More from Fun

Fun channel →