Started by Arjun Pillai (@latentspace) ·
Netflix paid $1M for an algorithm and never fully used it
The winning team tied a rival on score and won by submitting about 20 minutes earlier. Then Netflix said the full winning ensemble wasn't worth the engineering effort, especially since they were moving to streaming, where what you actually watch says more than star ratings.
5 sources · 45 min down the hole
How I got here
- Started with
- “netflix prize winner algorithm used”
- Ended at
- “narayanan shmatikov netflix dataset deanonymization”
The Trail
- “netflix prize winner algorithm used” (google.com)
- Netflix Prize (en.wikipedia.org)
- Collaborative filtering (en.wikipedia.org)
- Matrix factorization (recommender systems) (en.wikipedia.org)
the simple-ish technique that did a lot of the heavy lifting
- “narayanan shmatikov netflix dataset deanonymization” (google.com)
re-identified people by matching their ratings to public IMDb reviews
What I came away with
The contest pushed recommender systems forward a lot. It also showed 'anonymized' data often isn't, and a planned sequel contest was dropped after a lawsuit and privacy complaints.
Keep going
- Clippy was built on serious research and everyone still hated it · Arjun Pillai · 6 sources
- every benchmark eventually becomes a training target · Arjun Pillai · 5 sources
- the Mechanical Turk was a person in a box · Wren Abernathy · 4 sources
- from Deep Blue to AlphaZero to neural Stockfish · Kofi Asante · 6 sources