← back
How Google DeepMind is researching the next Frontier of AI for Gemini — Raia Hadsell, VP of Research
Original: How Google DeepMind is researching the next Frontier of AI for Gemini — Raia Hadsell, VP of Research
Takeaway
Frontier AI progress isn't just larger LLMs—omnimodal embeddings and probabilistic graph nets are quietly setting new SOTA in retrieval and weather forecasting.
Summary
- DeepMind VP of Research Raia Hadsell tours frontier directions beyond LLMs: Gemini Embeddings 2 (omnimodal, 8K text, 128s video, 80s audio, Matryoshka MRL).
- GraphCast and GenCast spherical/mesh graph neural networks for weather: 15-day global forecasts, 9-day hurricane landfall accuracy beating physics models by 3 days; GenCast outperformed 1300 gold-standard forecasts 97% of the time.
- Discusses Jennifer-Aniston-cell analogy motivating embedding models and DeepMind's root-node research strategy targeting deep, downstream-enabling problems.
- Frames DeepMind's mission as responsibly building AI for humanity across artificial, human, and robotic intelligence.
geminiembeddingsweather-forecasting
Original description
In this presentation, Raia Hadsell, VP of Research at Google DeepMind and AI Ambassador for the United Kingdom, opens AIE Europe and explores what's open in Frontier AI and the future of intelligence by focusing on advancements beyond standard large language models. She categorizes these innovations into three key areas: 00:00 Introduction 05:05 Advanced Embedding Models: Raia discusses the importance of embedding models for fast retrieval and recognition, similar to how the human brain uses 'Jennifer Aniston cells' to identify concepts across modalities. She highlights Gemini Embeddings 2, a fully omnimodal model that processes text, video, and audio into unified semantic vectors. 09:53 AI for Weather Forecasting: The team has developed revolutionary models for atmospheric prediction, moving away from traditional physics simulations. Notable breakthroughs include: 11:00 GraphCast: A spherical graph neural network that provides accurate 15-day weather forecasts. 12:47 GenCast: A probabilistic model that offers higher efficiency and accuracy (97% of the time compared to gold-standard benchmarks). 13:51 FGN: A functional generative network that directly predicts cyclone behavior, which is currently being utilized by the US National Hurricane Center. 14:35 World Models: Hadsell introduces Genie, a project focused on creating interactive, real-time environments. Starting from Genie 1 (2D platformers) and progressing to Genie 3, these models allow users to create and interact with high-quality, 3D photorealistic worlds. These environments demonstrate capabilities like memory, consistency, and the ability to be dynamically prompted by the user to change the surroundings in real-time. Speaker info: - https://uk.linkedin.com/in/raia-hadsell-35400266 - https://github.com/raiah