Meta's AI Guru Says AI Models Aren't Truly Intelligent!
Chief scientist of Meta, Yann LeCun, stated this at the AI Action Summit in Paris.
Speaking at the AI Action Summit in Paris, the chief scientist of Meta, Yann LeCun, stated that today’s artificial intelligence (AI) language models, like AI chatbots, are not as intelligent as humans. He explained that these models don’t have four major abilities that humans have.
Human-Like Actions
He said that these AI chatbots cannot remember information over time, provide logical reasoning, understand how the physical world operates, and plan complex tasks in a step-by-step manner. He further added that right now, these AI systems mostly rely on data to recognize patterns, as they are not yet able to truly understand the world.
To fix this, Meta is working on new ideas, and the firm is going to take new initiatives soon. As per the reports, the firm is building a world-based AI model. This AI model will incorporate retrieval-augmented generation (RAG), a feature that enables it to pull additional information from external sources, resulting in better and more accurate responses. Another is V-JEPA, a system that learns by guessing missing parts of the videos.
These innovations will help the AI to understand real-world situations in a better way. LeCun further stated that he believes that this type of AI model is the future of AI. According to him, these models will learn from real-life scenarios and will be able to think more abstractly, like humans do, to predict what happens next in complex situations. These improvements could make AI smarter and more capable of handling real-world challenges.
AI Improvements
Meta, a major tech company, is facing problems with its AI development. Many of the skilled researchers who built its 2023 AI model, Llama, have left. Out of the original 14 researchers, only three are still at Meta. Some researchers have joined a new company in Paris called Mistral. Mistral is moving quickly and gaining attention for its innovative work.
Meanwhile, Meta’s latest AI model, Llama 4, has failed to catch the developers’ attention and has not been able to impress them. Hence, they prefer other Firms’ AI chatbots, including OpenAI’s GPT-4o, Google’s Gemini 2.5 Pro, and Anthropic’s Claude 4 Sonnet, known for their superior reasoning abilities.
As of now, the firm has delayed the release of its latest and biggest new AI model, Llama 4, Behemoth. This initiative is making people wonder whether the firm will be able to keep up in the fast-moving AI world. As Meta works to improve intelligence with new “world-based” AI models, the company must retain its talented team and develop cutting-edge technology to remain competitive.
