Connor Leahy on a Promising Breakthrough in AI Alignment скачать в хорошем качестве

Connor Leahy on a Promising Breakthrough in AI Alignment

Connor Leahy is one of the world's leading figures in the field of artificial intelligence (AI), known for his contributions to the understanding and development of large language models (LLMs) and his advocacy for ethical AI practices. As a co-founder of Conjecture, an organization focused on AI safety and research, Leahy emphasizes the importance of creating AI systems that are aligned with human values and can be trusted to act in humanity's best interests. His work often explores the challenges of AI alignment, the risks associated with advanced AI technologies, and the need for collaborative efforts in the AI community to ensure the safe and beneficial development of AI. Leahy is recognized for his insights into the complexities of AI ethics, the potential existential risks posed by uncontrolled AI, and the importance of rigorous research and dialogue in mitigating these risks. Time stamp: 00:00 -- Introductory sequence 00:28 -- Introduction to the talk by Razo 02:51 -- Start of Interview 04:20 -- Leahy gives an overview of his work in AI alignment 10:05 -- Leahy talks about people who don't feel an AI threat 15:20 -- Leahy elaborates on intelligence and goals 19:55 -- Leahy on his approach to quantum physics 21:50 -- Razo on what should be the favored interpretation in science 26:04 -- Leahy on how to approach differences of opinion 38:12 -- Is Leahy optimistic about humanity's prospects? 43:50 -- Three potential scenarios for the future of humanity 48:10 -- Leahy on voting, mechanism design, and quadratic voting. 55:40 -- Leahy on what to do about human disagreement 1:01:26 -- How does Leahy reconcile his pessimism and his optimism? 1:04:35 -- Leahy on large language models and agent-based simulation 1:07:55 -- The two-state vector approach to science 1:09:15 -- Leahy on the importance of talking to politicians 1:13:55 -- Changing the assumption of 1-person 1-vote democracy 1:15:17 -- Leahy talk about the promising breakthroughs of Conjecture 1:21:55 -- Leahy on controlling AI but not fixing humanity 1:24:35 -- Where and how to follow Leahy