Subvolts take Google Gemini co-lead Oriol Vinyals discusses world models and the path to AGI. May 22, 2026 - Developers
- Learn about Google's research direction for advanced AI agents and memory systems.
- Business
- Understand Google's long-term AI strategy and potential breakthroughs in areas like robotics.
Features a leader from Google's Gemini program discussing core research for AGI and recent announcements.
Ep 87: Gemini Co-Lead on World Models, RL's Next Domains & Continual Learning
Audio streams from by Redpoint Ventures. Your place is remembered on this device.
From Subvolts
Summary and analysis by Subvolts editorial, AI-assisted.
Unsupervised Learning episode 87 features Oriol Vinyals, co-lead of Google's Gemini program. They discuss the research behind Google's AI announcements, focusing on world models as a path to AGI. The conversation covers agents, memory in AI, and the future of reinforcement learning, touching on robotics and scientific discovery.
This episode explores how Google is trying to build smarter AI using "world models" that understand how things work, like a child learning. They discuss how this approach could lead to AI that can do many complex tasks, potentially even scientific breakthroughs.
This episode offers insights into Google DeepMind's AI strategy, particularly their focus on world models for AGI. It's for AI researchers and enthusiasts interested in the practicalities of building advanced AI systems, including memory mechanics and the future of reinforcement learning. The discussion connects directly to current frontier lab efforts in developing more capable and general AI.
- HostJacob Effron
- HostJordan Segall
- GuestOriol VinyalsVP of Research, Google DeepMind
- Guest: Oriol Vinyals, VP of Research at Google DeepMind and co-lead of the Gemini program
- Show: Unsupervised Learning
- Episode: Ep 87: Gemini Co-Lead on World Models, RL's Next Domains & Continual Learning
- Length: 60 minutes
- Key Topic: World models as a path to AGI
- Related Concept: Spark release for AI agents
Oriol Vinyals, VP of Research at Google DeepMind and co-lead of the Gemini program, joins Jacob the day after Google I/O to unpack the research underpinning Google's latest announcements and where frontier AI is heading. The conversation moves from world models (why Google has uniquely bet on them as a path to AGI, what the "GPT moment" for video and images would look like, and how they connect to robotics and simulation) to agents (the Spark release, why the system and model need to be optimized jointly, and why scaffolding will eventually be written by models themselves). Oriol gets into the mechanics of memory in models, drawing on his cognitive neuroscience background to argue that file-system-style non-parametric memory is more practical than baking memory into weights at serving scale. He shares his views on the limits of RL today (LLMs are data-limited in a way that game-playing RL never was), why training on narrow domains like math and code generalizes surprisingly well, and what a true "Move 37" moment for science or ML research would look like. Throughout, he reflects on the unique advantages of being inside Google (TPU co-design, end-to-end revenue stability, the merger of Brain and DeepMind), the trade-offs between focus and exploration in research orgs, and why he believes AGI in some meaningful sense may already be here, even if the goalposts keep moving.
(0:00) Intro
(1:36) Why World Models
(4:21) The GPT Moment for Video
(7:51) What Makes Omni a World Model
(10:04) World Models & Robotics
(12:37) Evaluating Physics in AI
(14:51) Consumer Agents & Spark
(18:39) Scaffolding & the Bitter Lesson
(22:06) Memory & Continual Learning
(26:54) Research Bets Inside Big Labs
(32:30) Post-Training RL is Greenfield
(35:57) What Real Intelligence Looks Like
(39:11) RL Generalization
(43:00) Advice for Founders
(46:40) Can AI Truly Innovate?
(49:48) Recursive Self-Improvement
(52:14) Quickfire
Jacob is an AI investor at Redpoint Ventures. He's led Redpoint's investments in companies like Abridge, Physical Intelligence & Legora. Follow Jacob on Twitter (@jacobeffron).
On Unsupervised Learning we probe the sharpest minds in AI in search for the truth about what's real today, what will be real in the future and what it all means for businesses and the world. If you're a builder, researcher or investor navigating the AI world, this podcast will help you deconstruct and understand the most important breakthroughs and see a clearer picture of reality. Subscribe to this show to stay up to date on our latest episodes.
- World Models
- Oriol Vinyals
- AI systems that learn to understand how the world works to predict outcomes and plan actions.
- AGI
- Oriol Vinyals
- Artificial General Intelligence, AI that can perform any intellectual task a human can.
- RL
- Oriol Vinyals
- Reinforcement Learning, a type of machine learning where agents learn by trial and error.
- non-parametric memory
- Oriol Vinyals
- Memory that is stored externally, like in a database, rather than being built into the AI's core programming.
Same guests, other shows
42 minGemini co-leads on project origins and what's next
WithJeff Dean, Koray Kavukcuoglu, Noam Shazeer, Oriol Vinyals
49 minGemini 2.0 and the Evolution of Agentic AI with Oriol Vinyals
WithOriol Vinyals
27 minLife is like a game
WithOriol Vinyals, Max Jaderberg, Raia Hadsell, David Silver
56 min
Questions people ask
- What are world models in AI?
- World models are AI systems that learn to understand how the world works, enabling them to predict outcomes and plan actions. Google sees them as a key path to achieving artificial general intelligence (AGI).
- What is the Spark release?
- Spark is a release related to AI agents, where the system and model need to be optimized jointly for better performance.
Page generated Oct 19, 2026. Summaries are Subvolts' own; the audio, artwork and notes belong to by Redpoint Ventures.