Ilya's NeurIPS 2014 seq2seq talk: the real origin of scaling neural networks
peterjliu · x · 2026-10-07
Tracing the history of scaling, peterjliu points to Ilya Sutskever's sequence-to-sequence talk at NeurIPS 2014 as the real start of scaling neural networks for language, quoting his famous line: "If you have a large big dataset and you train a very big neural network, then success is guaranteed!"
More from Fun
- Fun trivia: When The Matrix was made, nobody knew matrix multiplication would underpin all AI — birchlse · 2026-10-07
- LLMs Assume Huge News Makes Front Pages, but Real Media Avoids It — QuintinPope5 · 2026-10-07
- Dean Ball proposes P = NP + AI as an equation that could shape the future — deanwball · 2026-10-07
- Developer Jokes About His AI Agent Missing Him During a Solo Movie Morning — threepointone · 2026-10-07
- "No windows, no dashboards" agent UI demo mocked for showing windows and dashboards — jasonkneen · 2026-10-07
- Teknium flags ChatGPT repeatedly recommending unknown third-party service, suspects ads — Teknium · 2026-10-07