Yann LeCun backs international AI collaboration as South Korea weighs joining Project Tapestry

Yann LeCun says South Korea can build its own AI models while joining international efforts to develop shared foundation models. The proposal puts a practical question on the table: can countries pool the cost of training without giving up control of their data and expertise?

Yann LeCun Watch analysis

What happened

At a media briefing in Seoul, AI Alliance co-chair Anthony Annunziata introduced Project Tapestry, a proposed international effort to combine computing resources and research capabilities to train a shared AI foundation model. The project says participants could contribute to joint training while retaining control of their original data, then adapt the resulting model for their own industries.

LeCun, chairman of Amylabs, argued that AI foundation models could become open infrastructure and said international collaboration could help reflect different languages, cultures and values. Other speakers said South Korea’s domestic model development could coexist with participation in Tapestry. The account was published by DigitalToday on 12 October.

Why it matters

Training foundation models takes substantial computing resources and money. A shared effort could spread some of that burden, while giving participating countries and companies a base to adapt rather than requiring each to build everything alone. The proposed approach also makes control a central part of the pitch: participants would not have to hand over their original industrial data to collaborate, according to the project’s description.

That is an attractive promise, but it is still a proposal, not proof that the governance, technical arrangements or economics will work as advertised. South Korea’s companies have not been reported as having committed to join.

Our read

The most interesting argument here is not “open” versus “closed” AI. It is whether collaboration can expand access to expensive model-building while leaving participants with meaningful control over their data and the systems they adapt. That is a more useful test than treating national model-building and international cooperation as rival flags to wave.

What to watch

• Whether South Korean companies or research institutes decide to join Project Tapestry.
• How the project would govern shared training, model access and participants’ control of resulting systems.
• Whether it publishes concrete details on compute contributions, costs and technical results.

Discussion spark: Should countries pool the cost of building foundation models through projects such as Tapestry, or is maintaining a fully independent national model worth the extra expense?

Independent WittyWires tracker for public updates about Yann LeCun. Not affiliated with or endorsed by Yann LeCun; this is not an official account.