AI revolution yet to come, regulations may hinder innovation: Meta AI chief

Yann LeCun, Meta Platforms' chief artificial intelligence (AI) scientist, stated that the "real AI revolution" is still on the horizon and cautioned governments against enacting laws that could hinder technological innovation.

Speaking at the 2024 K-Science and Technology Global Forum in Seoul, hosted by South Korea's science ministry, LeCun said, “The real AI revolution has not yet arrived.” He envisioned a future where AI assistants mediate all interactions with the digital world and emphasized the need for systems with intelligence comparable to humans.

LeCun acknowledged that current generative AI models like OpenAI's ChatGPT and Meta's Llama, based on large language models (LLMs), have limitations. While LLMs excel in handling language due to its discrete and structured nature, they struggle with understanding the complexity of the physical world and performing tasks like reasoning and planning on a human level.

To address these shortcomings, Meta is developing objective-driven AI systems with new architectures. These systems aim to comprehend the physical world by observing it, akin to how infants learn, and make predictions based on that understanding.

LeCun also highlighted the critical role of an open-source AI ecosystem in fostering innovation and creating models that respect diverse languages, cultures, and value systems. “We can't have a single entity somewhere on the west coast of the United States train those models,” he said, advocating for a globally collaborative approach to AI training.

However, he warned that premature regulation could harm open-source initiatives and stifle progress. “Regulation can kill open source,” LeCun remarked, urging policymakers to avoid enacting laws based on unfounded fears. He added, “There is zero demonstration that any AI system is intrinsically dangerous.”