Is LLM leading to AGI? Feifei Li, Yann LeCun vs Hinton, Ilya Sutskever

Over the past year or so, big language models have ignited a new wave of optimism about AI. There’s been a lot of debate about whether these models are truly conscious or if they can build a “world model.” The camps are divided. One side, the opponents, believe that big language models are just parrots, not understanding what they’re saying and lacking any genuine creativity. They see attempts to use these models to achieve artificial general intelligence as a misguided path. Yann LeCun and Fei-Fei Li are prominent voices in this camp.

Even though LeCun is Meta AI’s Chief Scientist and Meta has released the Llama series of large language models, he has repeatedly expressed disdain for LLMs. He believes that LLMs are essentially just autoregressive text generation models that have only a very superficial understanding of the world. They capture only the statistical patterns of text and can’t truly understand the meaning of text in the real world. He has even publicly stated that GPT will be abandoned in 5 years because its technology is not even capable of achieving the intelligence of a cat.