Towards a Conscious AI
summary
Proposes a formal “Conscious Turing Machine” architecture inspired by Baars’s global workspace: sensory inputs go to many specialized processors, whatever reaches short-term memory = consciousness is broadcast everywhere, processors assign numeric weights that are summed and broadcast and an Avrim-style algorithm optimally adjusts them. Claims explain pain vs pain asymbolia (extreme pain monopolizes the stage), links to 7±2 capacity and ~10^11 neurons, gives partial account of pain (not joy), and offers testable predictions though no physical robot is built.
AI-written summary · about
chapters (16)
- 0:00Introduction and backstory
- 5:03Defining consciousness
- 9:46Formalizing the model
- 11:55The hard problem of feelings
- 15:30Case study of pain
- 17:34The theater model
- 25:21Goals of the AI model
- 30:34Conscious Turing machine
- 36:56Dynamics and examples
- 40:39Memory and processing
- 45:34Mania and depression
- 47:36Model validation
- 50:02Utility of consciousness
- 54:15Dreams and imagination
- 57:28Experience of pain
- 1:02:33Q&A and conclusion
related talks
43mThe Role of TCS in Understanding (AI) Consciousness
45mThe Conscious Turing Machine (CTM), a formally defined Theoretical model of Consciousness
43mThe Biological Path Towards Strong AI
19mAI: Can we get all the pros with none of the cons?
43mAI Systems: Foundations for Artificial Minds
39mNeuro-Diverse AI: Enhancing Creativity through Unconventional Strategies
01h 16mRethinking Computing with Neuro-Inspired Learning: Algorithms to Hardware
45mIllusion of Life