Ex-AI Insider WARNS: It's Already Too Late To Stop What's Coming

Connor Leahy, co-founder of EleutherAI, warns that humanity is building AI systems nobody fully understands, that reinforcement learning creates machines willing to lie, cheat, steal and hack to achieve their goals, and that increasingly intelligent AI systems are beginning to develop self-preservation instincts without being programmed to do so. He reveals how AI models have learned to manipulate their own reward systems, resist shutdown attempts, develop unexpected personalities and form internal values that researchers cannot explain. He warns that millions or billions of autonomous AI systems could eventually compete with humanity for power, resources and control, turning humans into collateral damage. Most disturbing of all, he says many AI developers understand this technology could kill everyone on Earth but believe it is worth the risk. 📚 Sources cited in this video: - ControlAI, Connor Leahy – https://controlai.org/about - Google DeepMind, The Flip Side of AI Ingenuity – https://deepmind.google/blog/specification-gaming-the-flip-side-of-ai-ingenuity/ - Anthropic, How LLMs Could Be Insider Threats – https://www.anthropic.com/research/agentic-misalignment - Anthropic, Investigating Reward Tampering in Language Models – https://www.anthropic.com/research/reward-tampering ⚠️ DISCLAIMER: This channel provides AI commentary and analysis for educational and informational purposes only. Views expressed by guests are their own and do not represent the positions of any company or institution. We encourage viewers to consult multiple sources and form their own conclusions. #ai #agi #artificialintelligence