Human Brain vs AI Brain: Can We Design a Truly Human-like AI?
summary
Panelists argue current models remain abstract and miss cellular diversity, noting a 0.3 mm3 rat fragment contains 31,000 neurons and 55 types that networks do not model, and the brain runs on 20 watts. They call AlphaFold “very large‑scale pattern matching” and warn it’s data‑dependent, contrast training times from months to three days to 30 hours for later AlphaGo versions, cite ChatGPT’s hundreds of millions in training costs, and flag ownership and dataset monopolies as the real risk.
AI-written summary · about
chapters (15)
- 0:00Human vs machine intelligence
- 3:22Neuroscience as bridge
- 9:08AI limits vs human brain
- 12:02Optimal vs maximum intelligence
- 19:08AI successes and failures
- 23:50AlphaFold vs biological intelligence
- 27:50Energy efficiency in intelligence
- 31:02Oxytocin, gut microbes, and decision-making
- 32:50Brain region diversity
- 37:09Neuroscience and AI parallels
- 41:58AI’s societal impact
- 44:34AI surpassing human intelligence
- 47:02AI control and implications
- 48:06Alternative paths to artificial general intelligence
- 51:36Q&A
related talks
31mMyth & Truth behind Artificial General Intelligence
21mOrganoid Computing – Can Nature Save AI?
04mFrom big data to human-level artificial intelligence
12mAugmenting Human Capabilities to New Dimensions
21mSentient AI – Can We Create Artificial Consciousness?
24mFrom Neural Networks to Digital Brains: The Next Leap in AI
01h 14mTissue vs silicone musings on the future of deep learning hardware and software
14mAI Is Built to Be Applied