Subvolts take AI needs neuroscience to understand how brains learn so much from so little. Dec 30, 2025 - Developers
- Learn about biological learning principles that could inspire more efficient AI models.
- Business
- Understand potential future directions in AI research that could lead to more capable systems.
This is a solid interview with a researcher from a frontier lab discussing fundamental AI challenges.
Adam Marblestone — AI is missing something fundamental about the brain
Audio streams from Dwarkesh Patel. Your place is remembered on this device.
From Subvolts
Summary and analysis by Subvolts editorial, AI-assisted.
Dwarkesh Patel interviews Adam Marblestone on the Dwarkesh Podcast about what AI can learn from neuroscience. They discuss how brains learn efficiently from minimal data, the role of reward functions, and how genomes encode abstract goals. Marblestone, formerly of Google Deepmind, shares insights from his work on brain-computer interfaces and quantum computing.
This episode explores how human brains learn incredibly well with very little information, unlike current AI. It discusses how the brain's internal goals, encoded by our genes, might hold the key to making AI learn more efficiently and understand abstract concepts.
This episode is for AI researchers and neuroscientists interested in bridging the gap between biological intelligence and artificial systems. It explores fundamental questions about learning efficiency and goal encoding, areas directly relevant to current AI development at labs like Google Gemini and Anthropic. Understanding how the brain achieves complex learning with limited resources could inform future AI architectures and training methodologies, potentially leading to more capable and efficient AI models.
- HostDwarkesh Patel
- GuestAdam MarblestoneCEO, Convergent Research
- Guest: Adam Marblestone
- Show: Dwarkesh Podcast
- Length: 110 minutes
- Guest's Past Role: Research scientist at Google Deepmind on their neuroscience team
- Key Discussion Point: How the brain learns so much from so little
Adam Marblestone is CEO of Convergent Research. He’s had a very interesting past life: he was a research scientist at Google Deepmind on their neuroscience team and has worked on everything from brain-computer interfaces to quantum computing to nanotech and even formal mathematics.
In this episode, we discuss how the brain learns so much from so little, what the AI field can learn from neuroscience, and the answer to Ilya’s question: how does the genome encode abstract reward functions? Turns out, they’re all the same question.
Watch on YouTube; read the transcript.
Sponsors
* Gemini 3 Pro recently helped me run an experiment to test multi-agent scaling: basically, if you have a fixed budget of compute, what is the optimal way to split it up across agents? Gemini was my colleague throughout the process — honestly, I couldn’t have investigated this question without it. Try Gemini 3 Pro today gemini.google.com
* Labelbox helps you train agents to do economically-valuable, real-world tasks. Labelbox’s network of subject-matter experts ensures you get hyper-realistic RL environments, and their custom tooling lets you generate the highest-quality training data possible from those environments. Learn more at labelbox.com/dwarkesh
To sponsor a future episode, visit dwarkesh.com/advertise.
Timestamps
(00:00:00) – The brain’s secret sauce is the reward functions, not the architecture
(00:22:20) – Amortized inference and what the genome actually stores
(00:42:42) – Model-based vs model-free RL in the brain
(00:50:31) – Is biological hardware a limitation or an advantage?
(01:03:59) – Why a map of the human brain is important
(01:23:28) – What value will automating math have?
(01:38:18) – Architecture of the brain
Further reading
Intro to Brain-Like-AGI Safety - Steven Byrnes’s theory of the learning vs steering subsystem; referenced throughout the episode.
A Brief History of Intelligence - Great book by Max Bennett on connections between neuroscience and AI
Adam’s blog, and Convergent Research’s blog on essential technologies.
A Tutorial on Energy-Based Learning by Yann LeCun
What Does It Mean to Understand a Neural Network? - Kording & Lillicrap
E11 Bio and their brain connectomics approach
Sam Gershman on what dopamine is doing in the brain
<
- Amortized inference
- Adam Marblestone
- A way of making complex calculations more efficient over time.
- Reward functions
- Adam Marblestone
- The goals or objectives that guide learning and decision-making.
1 h 371 h 201 h 37AI researchers debate how close we are to recursive self-improvement
WithJohn Schulman, Beren Millidge, Charlie O’Neill
2 h 21
2 h 53
1 h 02Training General Robots for Any Task: Physical Intelligence’s Karol Hausman and Tobi Springenberg
WithKarol Hausman, Tobi Springenberg
1 h 5433 minAI Investor Panel: How Will We Fund the Global AI Revolution? | EP 219
WithAnjney Midha, Bonnie Chan, Dave Blundin
Questions people ask
- What is the main question discussed regarding AI and the brain?
- The episode discusses how the brain learns so much from so little data and what the AI field can learn from neuroscience.
- What is Adam Marblestone's background?
- Adam Marblestone is CEO of Convergent Research and previously worked as a research scientist at Google Deepmind on their neuroscience team.
Page generated Oct 19, 2026. Summaries are Subvolts' own; the audio, artwork and notes belong to Dwarkesh Patel.