events

From Word2Vec to Attention: The Idea Behind Every LLM

AI Office Hours continues its series on how language models work, moving from static Word2Vec vectors to contextual representations. The session covers why one fixed vector per word fails for a word like "bank", how meaning shifts with surr

Starts

2026-10-24T18:00:00+00:00

Organizer

AI Office Hours

Location

Online event

AI Office Hours continues its series on how language models work, moving from static Word2Vec vectors to contextual representations. The session covers why one fixed vector per word fails for a word like "bank", how meaning shifts with surrounding words, and how attention lets each word draw on its neighbours — illustrated by the classic puzzle over what "it" refers to. No heavy maths is required; the talk is beginner-friendly, with a recap for those who missed the earlier Word2Vec session and an open question-and-answer period.