Large language models (LLM), on which ChatGPT and other generative (content-creating) artificial intelligence (AI) tools are based, will never reach the level of human intelligence, believes Jan Lekun, chief scientist of the American Meta in the field of AI.
In his opinion, such models have a very limited understanding of logic and will not be able to reason and plan like a human. In addition, they “do not understand the physical world, do not have permanent memory, are unable to reason in any reasonable sense of the word, and cannot plan … hierarchically,” he said in an interview with the Financial Times.
Lecun argued against relying on LLM development in an effort to create human-level intelligence, as these models can only accurately respond to queries if trained on correct data, and are therefore insecure.
Instead, the expert is developing an entirely new generation of AI systems that he hopes will endow machines with “human” intelligence. However, this task could take a decade to complete, he explained.
Lecun leads a team of about 500 people at Meta’s basic AI research lab. They are working on creating an artificial intelligence that can develop a sense of common sense and study the structure of the world in a similar way to humans. This approach is called “world modeling.”
