The hype around artificial general intelligence (AGI) is pervasive, with tech luminaries like Nvidia’s Jensen Huang, Ben Goertzel, and Elon Musk making bold predictions about its imminent arrival.
Huang forecasts AGI within five years, Goertzel within three, and Musk as soon as the end of 2025. However, a dissenting voice comes from a prominent figure: Yann LeCun, Meta’s chief AI scientist and a Turing Award winner, is a notable sceptic.
LeCun argues that true AGI remains elusive because human intelligence itself isn’t general. He points out that current AI systems fall short in key areas like reasoning, planning, persistent memory, and understanding the physical world. These deficiencies limit AI applications like self-driving cars and smart assistants, making them error-prone.
LeCun is particularly critical of large language models (LLMs) like GPT-3 and Bard. He argues they have a superficial understanding because they rely solely on text, unlike humans who learn through real-world interaction.
LeCun proposes an alternative approach called “objective-driven AI.” This focuses on achieving specific goals set by humans. Instead of just text, these AI systems learn from the physical world through sensors and video data. This allows them to develop a comprehensive world model and predict the outcomes of actions. For example, the AI can plan and execute tasks more effectively by understanding how moving a chair affects a room’s layout.
LeCun believes that while machines will eventually surpass human intelligence, it will take time and won’t happen as soon as some predict. His view contrasts sharply with the more optimistic timelines suggested by other tech leaders. He emphasizes a more cautious and gradual approach to achieving advanced AI capabilities.
Based on: https://thenextweb.com/news/meta-yann-lecun-ai-behind-human-intelligence
Comments are closed.