*in-person* Anthony Hu | GAIA-1: A Generative World Model for Autonomous Driving

*Note* This talk will not be recorded. If you would like to see the talk, please RSVP. RSVPs will close on Wednesday 12th February at noon, or earlier if we reach capacity.

Title: GAIA-1: A Generative World Model for Autonomous Driving
Speaker: Anthony Hu (Applied Scientist, Wayve)
Paper: https://arxiv.org/abs/2309.17080
Abstract: Autonomous driving promises transformative improvements to transportation, but building systems capable of safely navigating the unstructured complexity of real-world scenarios remains challenging. A critical problem lies in effectively predicting the various potential outcomes that may emerge in response to the vehicle’s actions as the world evolves.
To address this challenge, we introduce GAIA-1 (‘Generative AI for Autonomy’), a generative world model that leverages video, text, and action inputs to generate realistic driving scenarios while offering fine-grained control over ego-vehicle
behavior and scene features. A world model is a predictive model of the future, allowing it to understand the consequences of its actions. We cast world modelling as a self-supervised sequence modelling problem, where the goal is to predict the next discrete token in the sequence. Similarly to language models, we show that the performance of world models scales gracefully with more parameters (~10B) and compute. Emerging properties of GAIA-1 include: generalisation to out-of-distribution states, contextual awareness, and understanding of 3D geometry.
Speaker Bio: I’m a researcher at Wayve, an autonomous driving company based in London, UK. My research interests are at the intersection of generative modelling, computer vision, and policy learning. I obtained a PhD from the University of Cambridge with my thesis entitled “Neural World Models for Computer Vision”. Prior to that, I went to Telecom Paris (a French Engineering Grande Ecole) and specialised in Maths and Computer Science.
Agenda:
5:15pm - Welcome desk opens (feel free to relax in the lobby before the talk)
6pm - Talk
6:40pm - Q&A session
7pm - Networking
8pm - Close
When RSVPing, please include your forename and surname for the venue's security.