My @TEDAI2024 talk is out! www.ted.com/talks/surya_... I discuss our work, spanning AI, physics, math & neuroscience, to develop a deeper scientific understanding of intelligence, both to understand biological intelligence and make better artificial intelligence. I focus on 5 gaps: - ThreadSky

suryaganguli.bsky.social • 80 days ago

My @TEDAI2024 talk is out!
https://www.ted.com/talks/surya_ganguli_can_ai_match_the_human_brain
I discuss our work, spanning AI, physics, math & neuroscience, to develop a deeper scientific understanding of intelligence, both to understand biological intelligence and make better artificial intelligence. I focus on 5 gaps:

Comments

suryaganguli.bsky.social•80 days ago

(1) Data efficiency: AI is ~ 10^5 more data hungry than humans. One approach to bridging this gap is more efficient pre-training: https://proceedings.neurips.cc/paper_files/paper/2022/hash/7b75da9b61eda40fa35453ee5d077df6-Abstract-Conference.html

suryaganguli.bsky.social•80 days ago

(2) Energy efficiency: AI is ~10^6 more energy hungry than humans. We study fundamental limits on accuracy as a function of energy: https://journals.aps.org/pre/abstract/10.1103/PhysRevE.108.014403
And show how the brain acts like a smart predictive energy grid:
https://www.nature.com/articles/s41586-021-03497-0

suryaganguli.bsky.social•80 days ago

(3) Going beyond evolution: We analyze novel quantum neuromorphic computers that implement neural algorithms in hardware where neurons are replaced by atoms and synapses are replaced by photons
https://pubs.aip.org/aip/apl/article/117/16/160501/1061343/Coherent-Ising-machines-Quantum-optics-and-neural
https://journals.aps.org/prx/abstract/10.1103/PhysRevX.11.021048
https://journals.aps.org/prx/abstract/10.1103/PhysRevX.14.011026

suryaganguli.bsky.social•80 days ago

(4) We show how to combine neuroscience and explainable AI to accelerate neuroscience discovery:
https://www.sciencedirect.com/science/article/pii/S0896627323004671

suryaganguli.bsky.social•80 days ago

(5) Melding minds and machines: We show how to meld minds and machines in bidirectional read-write loops to learn the language of the brain and then speak directly back to it:
https://www.nature.com/articles/s41586-022-04724-y
https://www.science.org/doi/10.1126/science.aaw5202

suryaganguli.bsky.social•80 days ago

Overall, I believe it is imperative to pursue a new unified Science of Intelligence that can help us understand and improve both biological and artificial systems alike. It must be done out in the open and will depend crucially on public investment in fundamental science.

rohitv.bsky.social•80 days ago

Incredible! Loved the problem formulation! Wow

Posting Rules

Be respectful to others
No spam or self-promotion
Stay on topic
Follow Bluesky's terms of service

Comments

Posting Rules

Reply