Embeddings for Everything: Search in the Neural Network Era

Dean's lecture, with Dan Gillick — Retrieval systems like internet search still use the same underlying keyword-based index they used back in the 1990s. Dan Gillick will describe his research on building a new kind of retrieval system based, somewhat unsurprisingly, on neural networks. He’ll try to explain the key pieces of technology and discuss how this may change the way we look for and find things.

Dan Gillick is a research scientist at Google and teaches machine learning and natural language processing in the MIDS program.

Видео Embeddings for Everything: Search in the Neural Network Era канала Berkeley School of Information