ChatGPT Will Not Achieve Human-Level Intelligence

Meta’s chief AI scientist, Yann LeCun, has made a bold statement regarding the limitations of large language models (LLMs) in reaching human intelligence. In an interview with the Financial Times, LeCun argued that LLMs like ChatGPT lack a true understanding of the physical world, persistent memory, reasoning abilities, and hierarchical planning skills.

LeCun cautioned against relying on LLMs to achieve human-level intelligence, citing their dependence on specific training data and inherent safety risks. Instead, he is focusing on developing a new generation of AI systems that aim to imbue machines with human-like intelligence, a goal that may take up to 10 years to accomplish.

This perspective comes at a time when other companies are pushing forward with enhanced LLMs in pursuit of artificial general intelligence (AGI), which surpasses human cognition. For instance, AI firm Scale recently raised $1 billion in funding, while French startup “H” secured $220 million, both with ambitions of achieving full AGI.

However, not all experts share the optimism surrounding AI’s potential to replicate human intelligence. Akli Adjaoute, a seasoned AI professional and author, views AI as a powerful tool with limitations in understanding and reasoning like humans. He emphasizes the importance of recognizing AI’s reliance on algorithms and patterns for specific tasks, rather than expecting it to think and reason like humans.

As the race for AGI continues, the debate over AI’s true capabilities and limitations remains a topic of interest and discussion among industry experts and investors.