

Next goal: 10
Nomination created on December 4, 2025
Champion Hub
6 people want this. Add your vote.
0 of 6 done
Champion Hub
Every way to move this nomination, in one place.

Neuroscientist at CSHL. Interests: neuroAI, molecular connectomics, & cortical circuits underlying decisions. Co-founder of Cosyne and NAISys meetings.

Dive into the cutting-edge worlds of technology and science with Dwarkesh Patel's meticulously researched interviews that bring together brilliant minds to unpack revolutionary ideas in AI, biotech, and beyond. Each episode delivers profound, thought-provoking conversations that challenge conventional wisdom and illuminate the paths shaping humanity's future.
Tony Zador is at the center of the biggest brain-mapping breakthrough of the past year, as a core member of the International Brain Laboratory and coauthor on the September 3, 2025 Nature paper that charted how prior expectations are represented across 600,000 recorded neurons in the mouse brain. That work, covered widely and highlighted by his lab, reframes decision making as a brain-wide computation and has immediate implications for how we build next-gen AI systems that reason under uncertainty.
He is also helping stand up the pipeline that will turn those insights into talent, co-directing Cold Spring Harbor Laboratory’s new BioAI Ph.D. program launched on December 4, 2025 to train researchers fluent in both neuroscience and machine learning. Add his group’s recent upgrades to BARseq, which can now map millions of neurons, and you have a guest who can connect fresh experimental results to concrete AI progress today.
There is no record of Zador previously appearing on the Dwarkesh Podcast, which makes this a timely first conversation for an audience obsessed with how biology can supercharge AI. Support this nomination now so listeners can hear the scientist shaping how brains and machines learn, while the field is moving fast.
Record a short video telling everyone why this dialogue should happen. Your clip gets featured right here.
No voices yet. Submit a tweet, video, or post to be the first.
Suggested
How much of human 'language' is really an abstract computational ability shared with animals?
You've studied how the cortex processes symbolic structure — whether it's a mouse's whiskers or a human's sentences. What evidence suggests the brain's core operation is a kind of language-of-thought, and could we test that by decoding non-linguistic species?
The connectome is too expensive to be the repository of all learned knowledge; maybe sparse connectivity itself is the underlying code
You've made the point that wiring is metabolically expensive and that synaptic connectivity likely reflects a compressed, structural memory. If that's true, does it suggest that artificial networks should explicitly learn connectivity patterns (not just weights) as the primary representation?
Evolution hardwired powerful priors into brain architecture; Transformers need something equivalent
Humans learn from a few examples because evolution pre-shaped the loss landscape. Modern AI grossly underuses structural priors like spatial hierarchy, sparse connectivity and predictive timing. What would a 'mutated transformer' look like if we tried to evolve its wiring instead of growing it in pruned runs?