events
From Word2Vec to Attention: The Idea Behind Every LLM
A beginner-friendly online session run by AI Office Hours that continues its series on word representations, moving from static Word2Vec vectors to contextual embeddings.

En anglais
- Starts
- 2026-10-24 18:00 UTC
- Organizer
- AI Office Hours
- Location
- Online event
A beginner-friendly online session run by AI Office Hours that continues its series on word representations, moving from static Word2Vec vectors to contextual embeddings. The talk sets out, without heavy maths, why one fixed vector per word is insufficient for words with several meanings, how contextualised embeddings address that, and how the attention mechanism lets each word draw meaning from its neighbours — illustrated with the classic sentence about an animal, a street and what "it" refers to. It closes with a short open Q&A, and a recap is offered for anyone who missed the earlier session.
Event website