110 Disturbing Facts About Artificial Superintelligence to Fall Asleep To

What happens when artificial superintelligence becomes smarter than humans—and how could we keep it safe? Explore 110 fascinating facts about AI safety, superintelligence, alignment, recursive self-improvement, interpretability, and AI governance. Discover how reward hacking, model theft, concentrated computing power, and unreliable oversight could shape the future—and why no single safeguard is enough. Settle in for a thought-provoking journey through the risks and possibilities of advanced AI. Subscribe for more fascinating videos to fall asleep to. Sources: • Superintelligence: Paths, Dangers, Strategies — Nick Bostrom (2014). • Human Compatible: Artificial Intelligence and the Problem of Control — Stuart Russell (2019). • Concrete Problems in AI Safety — Dario Amodei et al. (2016). • Risks from Learned Optimization in Advanced Machine Learning Systems — Evan Hubinger et al. (2019). • On the Opportunities and Risks of Foundation Models — Rishi Bommasani et al. (2021). 00:00 Introduction 03:51 1 - 7: Defining Superintelligence 15:03 8 - 14: Goals, Consciousness, and Side Effects 26:39 15 - 21: Misaligned Rewards and Human Values 37:55 22 - 28: Reliability and Evaluation Failures 49:13 29 - 35: Limits of Interpretability 1:00:23 36 - 42: Explanations and Oversight Limits 1:11:21 43 - 49: Alignment Through Debate and Monitoring 1:21:26 50 - 56: Attacks, Backdoors, and Model Theft 1:31:36 57 - 63: Data, Compute, and Model Control 1:41:40 64 - 70: Power, Access, and Accountability 1:52:15 71 - 77: Social Costs of AI Automation 2:03:05 78 - 84: AI Governance and Shared Benefits 2:13:21 85 - 91: Oversight and Accountability 2:24:26 92 - 98: Limits of Scaling Forecasts 2:34:36 99 - 105: Physical and Predictive Limits 2:44:51 106 - 110: Limits of Intelligence and Safeguards 2:52:50 Conclusion Thanks for watching