Yann LeCun, a key figure in the world of artificial intelligence and known as one of the “Godfathers of AI,” has expressed his wish for AI models to expand their capabilities beyond just text generation. In a recent post on X (formerly Twitter), LeCun shared his thoughts while responding to entrepreneur Paul Graham. Graham had mentioned that if he were 17 years old today, he would focus on building and training large language models (LLMs) with all the resources available to him.
LeCun responded by stating, “I’d want to know why LLMs can help me write essays but can’t handle cleaning my room. I’d then focus on studying subjects that could address that issue.” He envisions developing new methods and systems that would enable AI to learn physical tasks as efficiently as humans and animals do. He emphasized that he would pursue this goal at any age.
In another exchange, LeCun reacted to comments from his former colleague at Meta, Jitendra Malik, who recently joined Amazon. Malik shared his thoughts on the importance of precise terminology in scientific discussions, especially regarding LLMs and visual models. He argued that mixing up these terminologies lacks clarity and can cause confusion. LeCun disagreed with Malik, insisting that the understanding of world models, which are essential for robotics, should not be conflated with models that focus solely on video generation or prediction.
For those unfamiliar, Yann LeCun is a prominent French-American computer scientist known for his groundbreaking work in deep learning, particularly for creating convolutional neural networks (CNNs). He has served as the Chief AI Scientist at Meta and is a professor at New York University. Recently, LeCun announced his decision to leave Meta after a 12-year tenure, where he spent five years as the founding director of the Facebook AI Research (FAIR) lab and seven years in his current role. He acknowledged the significant impact of FAIR on the field of AI and the tech community during his time there.
