Meta’s AI Chief Explains Why LLMs Cannot Attain Human-Like Intelligence

Meta’s AI chief Yann LeCun has made a bold statement regarding the limitations of current AI chatbots like ChatGPT, saying that they will never be able to achieve human-level intelligence due to their lack of reasoning abilities. This comes at a time when tech giants are racing to develop Artificial General Intelligence (AGI) – a model that can match or surpass human intelligence.

LeCun emphasized that companies should not rely on current Large Language Models (LLMs) to reach human-level intelligence, as these models lack the ability to understand the physical world, have persistent memory, reason, or plan hierarchically. He described them as ‘intrinsically unsafe’ due to their reliance on specific training data to answer human prompts accurately.

However, LeCun remains optimistic about the future of AI, stating that a new generation of AI systems based on ‘world modeling’ could potentially achieve human-like intelligence within the next decade. His team at Meta’s Fundamental AI Research lab is actively working towards this goal.

LeCun also highlighted the efficiency of animal intelligence compared to current AI systems, noting that animals and humans can become highly intelligent with significantly less training data. He pointed out that even corvids, parrots, dogs, and octopuses can surpass the capabilities of current AI models with their limited neural networks and parameters.

As the race for AGI continues, LeCun’s insights shed light on the challenges and possibilities of AI development, emphasizing the need for innovative approaches to achieve human-level intelligence in machines.