How Google DeepMind is researching the next Frontier of AI for Gemini — Raia Hadsell, VP of Research

By ai.engineer

Categories: AI, Tools

Summary

Google DeepMind's VP of Research reveals that embedding models—inspired by neuroscience's Jennifer Aniston cells—are critical companions to generative AI, enabling fast retrieval and cross-modal understanding. The newly released Gemini Embeddings 2 is fully omnimodal, representing the frontier of AI beyond language models.

Key Takeaways

1. Embedding models encode concepts using contrastive losses to achieve fast retrieval, recognition, and comparison across modalities—solving the same problem brains use Jennifer Aniston cells for, where neurons activate identically across images, video, and audio.
2. Gemini Embeddings 2 is fully omnimodal and derived from Gemini, providing knowledge and understanding needed for robust cross-modal representation—a critical infrastructure layer builders need alongside generative models.
3. DeepMind's research strategy prioritizes finding 'root nodes'—deeply unsolved fundamental problems—rather than optimizing leaves, enabling downstream impact across entire problem spaces.
4. Building frontier AI requires thinking beyond artificial intelligence to include human intelligence augmentation and robotics intelligence simultaneously, not in isolation.
5. Raia Hadsell manages ~1,200 scientists and engineers across 10 DeepMind labs working on distinct frontier areas—illustrating the scale of infrastructure needed for frontier AI research at major tech companies.

Topics

• Embedding Models
• Omnimodal AI
• Contrastive Learning
• Frontier AI Architecture
• Gemini Embeddings

Transcript Excerpt

Our next speaker is VP of research at Google DeepMind. Please join me in welcoming to the stage Rya Hazdell. Hello everyone. Wonderful. Uh what a lovely full room and good smiles. I heard the uh dig on Google there at the end. I I I I did catch that. Um so my name is Rya Hadzel. Uh I've been a part of uh DeepMind uh for the last uh almost 13 years and I'm very happy to have AI engineer come here to London. I'm also uh very proud this year to be a UK AI ambassador. So I help the government um aca...