‘Open-source will win’: Meta’s Yann LeCun takes aim at competitors’ closed AI models

Yann LeCun, Meta’s Chief AI Scientist, didn’t hold back when discussing Meta’s open-source approach to AI at Nasscom AI Confluence in Delhi on Thursday, making pointed comparisons to competitors like OpenAI and Google. LeCun, a pioneer in AI, suggested that Meta’s commitment to transparency and collaboration gives it an edge over rivals who opt for more closed-off models, especially when it comes to Large Language Models (LLMs).

LeCun recounted Meta’s decision to open-source its models like LLaMA, contrasting it with OpenAI’s approach. “When OpenAI produced GPT-3, we produced something called OPT-3, which was a similarly powerful model, but it was completely open-source,” he said, highlighting how Meta’s transparency allowed researchers worldwide to build upon their work. “OpenAI never released GPT-3 in open source,” he added, implying that keeping such models closed restricts innovation and collaboration.

Meta’s choice to keep its models accessible is based on the belief that open-source platforms are inherently more powerful and secure than proprietary systems. LeCun made a strong case for the superiority of open-source approaches in the AI landscape: “We know for a fact that open-source software platforms are both more powerful and more secure than the closed-source versions.” This, according to LeCun, is because more people can scrutinise and improve the technology, leading to faster advancements.

The competition between Meta and other AI giants isn’t limited to models alone. LeCun also noted the differences in how Meta and its competitors view the risks associated with LLMs. “Three weeks after we took down Galactica, ChatGPT was released, and that was the second coming of the Messiah,” he quipped, referring to the hype surrounding OpenAI’s ChatGPT. LeCun hinted that the public’s response to ChatGPT might have been overblown compared to the actual risks these technologies pose, adding, “We also concluded that because the world didn’t end after ChatGPT, the dangers of the risk associated with those models were basically overblown.”

LeCun’s remarks underscore Meta’s confidence in its strategy of embracing open-source models and its long-term commitment to advancing AI through collaboration. As other companies continue to lock their systems behind closed doors, Meta is betting that openness will ultimately lead to superior and more secure AI solutions.