Myth & Truth behind Artificial General Intelligence

on YouTube ↗31m

" summary

Claim: current deep learning on GPUs and TPUs cannot produce AGI because it is statistical and power-hungry; building AGI requires neuromorphic processors that mimic brain memory+compute and spiking signals. Evidence: brain-scale is about 100 billion neurons running at ~30 watts vs megawatts for large datacenters. Existing platforms like Spinnaker and BrainScaleS reach roughly mouse-scale (orders of magnitude below human) but use far less energy. Projection: with scaling of neuromorphic hardware and software, human-level AGI plausibly appears within decades.

AI-written summary · about

" chapters (28)

1. 0:00AGI myths and truths
2. 1:09General vs narrow intelligence
3. 2:33Why narrow AI falls short
4. 3:38Gardner’s nine intelligences
5. 5:03Measuring intelligence progress
6. 6:27Path to AGI is multidisciplinary
7. 6:48Biological vs non-biological intelligence
8. 8:58Hierarchy in biological systems
9. 11:28AGI hardware options
10. 13:16Computing power evolution
11. 14:41Moore’s Law limits
12. 15:43Energy gap: brains vs computers
13. 16:27Raw compute can’t achieve AGI
14. 17:12Neuromorphic computing
15. 18:39Neuromorphic hardware limits
16. 20:05Neuromorphic scaling potential
17. 21:06Analog vs digital neuromorphic
18. 22:31Quantum computing’s role
19. 23:13Digital vs neuromorphic vs quantum
20. 24:17Physics and neuroscience for AGI
21. 25:21Objective vs subjective AGI
22. 26:04AGI theories hardware data
23. 26:49Scaling to brain size
24. 27:32Global AGI investments
25. 27:52AGI risks and should we build
26. 28:34AGI coexistence scenarios
27. 29:57Why deep learning fails
28. 30:39AGI timeline and takeaways

" related talks

• 45mThe Curious Incident of Developing Artificial General IntelligenceJimmy Ba · Unknown Futures of Generalization 2024
• 01h 19mThe Al Revolution: The Road to SuperintelligenceTim Urban · Øredev 2015
• 30mBeyond GPUs: Why AI Agents Need a Need New HardwareVasanth Mohan · AI By the Bay 2025
• 52mHuman Brain vs AI Brain: Can We Design a Truly Human-like AI?S. R., D. O., J. V. · DSC Europe 24
• 09mSmale’s 18th Problem and Intelligence Science 2Michael Xuan, Brian Athey · Smale@95: A Conference in Honor of Steve Smale
• 51mBeyond Deep Learning: Next Generation AIDanko Nikolic · DSC DACH 2023
• 13mArtificial General Intelligence: Why aren't we there yet?Gary Marcus · Collision 2017
• 25mThe Great AI GameAzeem Azhar, Carl Benedikt Frey · DLD Summer 2021