在周二于Meta伦敦举办的一次活动中，图灵奖得主、Meta AI部门负责人Yann LeCun对通用人工智能（AGI）的概念表示了深度的怀疑。他认为，人类智能并非普遍适用，因此追求“人类水平的AI”更为实际。LeCun指出，当前的AI系统在推理、规划、持久记忆和理解物理世界等四个关键认知领域存在显著局限，这限制了它们的性能并可能导致错误。
他以自动驾驶汽车和家用机器人为例，强调了这些AI应用在应对复杂现实世界情境时的不足。尽管大型语言模型（LLMs）在语言生成方面表现出色，但LeCun认为它们对现实世界的理解极其有限，主要依赖于文本数据，而人类则通过全方位的感官体验来学习和理解世界。他甚至估计，一个四岁的孩子在成长过程中接触的数据量可能是最大LLMs的50倍。
LeCun的观点强调了AI研究的挑战，即如何让机器具备更接近人类的交互和学习能力。他的这些言论进一步引发了关于AI发展方向的讨论，即是否应当更加专注于发展具有特定领域专业知识的AI，而非追求全面的AGI。
英语如下：
News Title: “Turing Award Winner Yann LeCun: Large Language Models Fall Short of Human Intelligence, AGI Dreams Fading?”
Keywords: Turing Award, Large Language Models, Human Intelligence
News Content: During an event at Meta’s London office on Tuesday, Yann LeCun, a Turing Award recipient and head of Meta’s AI division, expressed deep skepticism about the concept of Artificial General Intelligence (AGI). He argued that human intelligence is not universally applicable, making the pursuit of “human-level AI” a more realistic goal. LeCun pointed out that current AI systems exhibit significant limitations in four key cognitive domains: reasoning, planning, long-term memory, and understanding the physical world, which hinder their performance and can lead to errors.
Using autonomous vehicles and household robots as examples, he emphasized the shortcomings of AI applications when dealing with complex real-world scenarios. While large language models (LLMs) excel in language generation, LeCun believes their understanding of the real world is extremely limited, primarily relying on textual data. In contrast, humans learn and comprehend the world through a holistic sensory experience. He estimates that a four-year-old child might be exposed to 50 times more data in their development than the largest LLMs.
LeCun’s perspective highlights the challenges in AI research, namely, how to enable machines to interact and learn more similarly to humans. His remarks have further fueled discussions about the direction of AI development, whether to focus more on domain-specific AI with expert knowledge rather than pursuing a comprehensive AGI.
【来源】https://thenextweb.com/news/meta-yann-lecun-ai-behind-human-intelligence
Views: 1