Yann LeCun, Meta’s chief AI scientist, recently argued that despite rapid progress, artificial general intelligence (AGI) on par with human cognition remains a distant goal.
LeCun cited the limitations of today’s dominant language models, trained on vast amounts of text data. He argues this narrow approach cannot instill common sense or general reasoning abilities that characterize human intelligence.
His sober view contrasts Nvidia CEO Jensen Huang’s claim that AI could match humans across various cognitive tasks in under 5 years. LeCun suggests the chipmaker has much to gain from fueling AGI hype and the ensuing hardware demands.
Instead, LeCun believes animal-level intelligence is a more achievable milestone in the coming years. Meta is pursuing multimodal AI that can understand connections across images, video, speech, and other data beyond text. Still, the fundamental breakthroughs needed for human-level AGI likely remain decades away.
LeCun is also skeptical about quantum computing, believing specialized AI hardware and neural processing chips hold more practical promise for continued AI progress. While cautioning against overinflated expectations, Meta remains committed to fundamental AI research for emerging general applications.
