Meta AI Chief Critiques Current AI Models, Announces New Initiatives
Yann LeCun, Meta’s chief scientist, delivered a stark assessment of contemporary AI systems at the AI Action Summit in Paris. Current language models, including popular chatbots, lack fundamental human-like intelligence, he argued. These systems cannot retain long-term memories, engage in logical reasoning, comprehend physical-world mechanics, or execute multi-step planning.
LeCun emphasized that today’s AI relies predominantly on pattern recognition from training data rather than genuine understanding. Meta plans to address these limitations through new research initiatives, including development of a "world-based" AI model incorporating retrieval-augmented generation techniques.