Tim O'Reilly: Big AI Labs Are Reading the Future Wrong

"The big labs are reading the future wrong." That's how Tim O'Reilly, the publisher whose book empire generative AI is busy scraping, sums up his read of the frontier labs in a new Wired interview with Steven Levy. The piece is making the rounds in our Who's Who directory, where two tracked experts have shared it.

O'Reilly's charge is specific. The labs, he says, "have told themselves a narrative where having the biggest, best model is the key to the future," and he argues that bet is misaligned with what people actually reach for. "Big models like Claude are optimizing for particular use cases, but they aren't necessarily the use cases that people want," he tells Levy, pointing to chatter that newer flagships like Fable and Sol are worse writers than earlier models.

His alternative is open source, but broader than the version that has taken hold in the AI discourse. Levy opened by asking him to make his case for being all in, and O'Reilly's answer starts by rejecting the frame: opening up neural-net weights is only a fraction of it. What he wants is unlocking the whole stack of an AI system, handing control back to designers and users rather than to whoever ships the biggest model.

The two do not agree on everything. In the same conversation Levy and O'Reilly land on a clear disagreement over whether AI can produce original creative content. O'Reilly, who treats AI as a new creative medium and even keeps a blog of his chats with it, is on the yes side.