Tim O'Reilly Urges Open-Source AI, Attacking Big Labs' Control Architecture

1 articles · Updated · WIRED · Aug 14

Tim O’Reilly argued AI should be opened beyond model weights to the full stack, saying users and developers need control over models, harnesses and applications.
Big labs are backing the wrong future, he said, by chasing ever-larger frontier models and building systems that lock in users, track behavior and limit customization.
O’Reilly contended open systems would spread AI more effectively through society, warning the US could still lose to China if lower-level models are adopted more widely there.
On safety, he said cybersecurity and pathogen risks argue for slowing frontier models rather than restricting open-weight models, which he cast as less central to recent incidents.
He tied that view to a broader critique of Silicon Valley, saying capital is concentrating AI power in a few firms even as the real innovation may come from open-source efforts outside VC funding.