Meta 首席科学家、图灵奖得主 Yann LeCun 近日在接受知名科技博主 Lex Fridman 的深度访谈时，明确表示他认为人工智能（AI）毁灭人类的可能性为零。这场长达近三个小时的对话，涵盖了从大型语言模型（LLM）的局限性、大模型的误解，到通用人工智能（AGI）面临的挑战，以及对 AI 末日论的深入批判。
LeCun 在访谈中指出，尽管 AI 技术发展迅速，但目前的 AI 系统仍然存在显著的局限性，远未达到能够自我意识或具备独立决策能力的水平。他强调，公众对 AI 的一些恐惧往往源于对技术的误解，即“大模型幻觉”，即认为模型越大，智能越高，而忽视了模型的训练目标和实际应用的局限。
在讨论 AGI 时，LeCun 表示，尽管 AGI 的实现将是科学上的重大突破，但目前的技术路径和理论基础还存在诸多挑战。他认为，AI 研究者应当更加专注于解决实际问题，而不是过分追求理论上的完美。
对于 AI 末日论，LeCun 表达了明确的反对态度，认为这种担忧被过分夸大了。他强调，AI 的风险更多来自于其在社会、经济和政策层面的应用，而不是技术本身具备的破坏性。因此，他呼吁科学家、政策制定者和公众共同关注如何在 AI 发展中确保伦理和安全，以实现科技与社会的和谐共进。
此次访谈为公众提供了一个深入了解 AI 的窗口，同时也揭示了 AI 研究者对于未来科技发展的深思熟虑和谨慎态度。来源：AppSo。
英语如下：
News Title: “Turing Award Winner Yann LeCun: AI Doomsday Odds Zero, In-Depth Interview Offers New Perspectives on AI”
Keywords: Yann LeCun, AI Safety, Chief Scientist at Meta
News Content: In a recent extensive interview with renowned tech podcaster Lex Fridman, Yann LeCun, Chief Scientist at Meta and Turing Award laureate, categorically stated that he believes there is zero chance of artificial intelligence (AI) destroying humanity. The nearly three-hour conversation delved into topics ranging from the limitations of large language models (LLMs), misunderstandings about big models, the challenges facing artificial general intelligence (AGI), and a critique of doomsday scenarios related to AI.
LeCun pointed out during the discussion that while AI technology is advancing rapidly, current AI systems still have significant limitations and are far from achieving self-awareness or independent decision-making capabilities. He emphasized that much of the public’s fear surrounding AI often stems from misunderstandings, a phenomenon he refers to as the “big model illusion,” which assumes that larger models equate to higher intelligence, overlooking the limitations of training objectives and practical applications.
Addressing AGI, LeCun acknowledged that its realization would be a major scientific breakthrough, but he highlighted that current technological pathways and theoretical foundations present numerous challenges. He advocated for AI researchers to focus more on solving real-world problems rather than solely pursuing theoretical perfection.
Regarding AI doomsday scenarios, LeCun expressed a firm counterposition, contending that such concerns are overblown. He stressed that the risks associated with AI are more likely to stem from its societal, economic, and policy applications, rather than the inherent destructiveness of the technology itself. Consequently, he called for scientists, policymakers, and the public to collaborate on ensuring ethics and safety in AI development to foster a harmonious coexistence between technology and society.
This interview offered the public a profound insight into AI and showcased the thoughtful and cautious approach of AI researchers toward the future of technological advancement. Source: AppSo.
【来源】https://mp.weixin.qq.com/s/ny6awp7BMvnRCZbn4lLTsA
Views: 1