A public clash between two popular AI researchers has again brought the limelight to the question about artificial intelligence- can machines ever match human intelligence, and what does that even mean? Both Google DeepMind CEO Demis Hassabis and Meta’s chief AI scientist Yann LeCun are at odds over the idea of general intelligence, a key concept behind artificial general intelligence, also known as AGI. Their disagreement has become a hot topic on X, getting widespread attention across the industry, even prompting a response from Tesla CEO Elon Musk.
AGI is popularly known to be the future stage of AI where systems can understand, learn and solve unfamiliar problems across multiple domains, much like humans do. While today’s AI models can outperform people in narrow tasks, from exams to complex games, most experts agree they still lack true adaptability and reasoning.
LeCun claims that general intelligence does not exist, even in humans. According to him, human intellect is highly specialised, influenced by biology, experience and situation. People succeed in various areas, yet cognitive limitations apply to all. He also cited chess as an example, claiming that even the best player can only consider a small number of options, whereas machines can calculate millions of moves rapidly.
This viewpoint has prompted LeCun to crticise current AGI initiatives, arguing that expanding data and computer capacity alone would not result in human-like intelligence. He has instead advocated for new ways that emphasise memory, thinking and enhanced sensory comprehension.
He later argued that LeCun is confusing general intelligence with limitless intelligence. While humans have limits, Hassabis stated that the human brain remains the most flexible learning system known, capable of adapting to an extraordinary range of tasks. He believes AI systems can eventually reach a similar level of generality.
On the other hand, Elon Musk stated that the DeepMind CEO was correct. Musk has been vocal about AI and said that highly advanced AI is inevitable, even as he warns about its risks. Despite the exchange, LeCun remains firm, clarifying that his objection is largely about language. Calling AI general, he says, wrongly implies a human-like intelligence that may not be a realistic or meaningful benchmark.