In a recent interview with the Financial Times, Meta’s AI chief Yann LeCun shared insights into the limitations of large language models (LLMs) like OpenAI’s ChatGPT and Google’s Gemini. LeCun stated that these AI systems will never be able to reach human levels of intelligence due to their limited understanding of logic, lack of persistent memory, inability to understand the physical world, and incapacity to plan hierarchically. He emphasized that these models are only accurate with the right training data, making them “intrinsically unsafe” and not suitable for achieving human-level AI.
LeCun and his team at Meta’s Fundamental AI Research lab are working on developing a new generation of AI systems based on a different approach called “world modeling.” This approach involves the system building an understanding of the world around it, similar to how humans do, and predicting what would happen if something changes based on this understanding. LeCun predicts that it could take up to 10 years to achieve human-level AI using this world modeling approach.
Meta, the parent company of popular tech platforms like Facebook, Instagram, and WhatsApp, is a leader in generative AI with its LLMs like Llama 3. However, the company faced a significant decline in market value in April after announcing plans to heavily invest in AI. CEO Mark Zuckerberg reassured investors to be patient as the company builds out its AI capabilities, comparing the strategy to the successful monetization of past products like Reels and Stories.
Despite the skepticism from investors, Zuckerberg’s net worth is estimated at $165.5 billion, making him one of the world’s wealthiest individuals. Along with his wife Priscilla Chan, Zuckerberg has committed to donating 99% of their stake in Meta within their lifetimes. The company’s focus on AI, even at significant expense, reflects its belief in the long-term potential of developing advanced AI systems.
With Meta’s continued investment in AI, particularly through the development of new world modeling systems, the company aims to differentiate itself from competitors like OpenAI and Google. LeCun’s insights into the limitations of current LLMs provide a glimpse into Meta’s approach to advancing AI technology and achieving human-level intelligence. As the tech giant navigates the challenges of investor confidence and market fluctuations, its commitment to innovation and research in AI remains a top priority.
