A keynote turns into a split

At Dreamforce's opening keynote, 12,000 attendees watched Nvidia's Jensen Huang argue for full speed on AI minutes after Anthropic's Dario Amodei made the case for slowing down.

The letter that started it

On Sept. 12, Amodei published a 3,800-word open letter laying out a plan to pace AI development. Sam Altman and Elon Musk agreed with parts of it, with Altman pushing for independent evaluators.

Huang: safety is engineering

Asked by Marc Benioff about pacing, Huang framed safety as a solvable build problem rather than a reason to hold back, and said new laws and regulations are unnecessary.

“Safety is an engineering problem. We're developing software after all.”

Let the market decide

Huang's alternative to regulation is company judgment: run flat out, and stop only if you think the product is unsafe. He called not shipping an unsafe product the obvious move.

“Run as fast as you can and if you feel the product's not going to be safe, take a pause.”

Amodei's three steps

Labs should audit their own safety record even without an incident, recommit to transparency, and spend more on safety. Anthropic has committed to step one and wants industry-wide, eventually international, standards.

“It's very tempting to attack your competitor and say 'these guys are unsafe.'”

Salesforce wants the time

A slowdown would help the host. Salesforce president Patrick Stokes said the company has catching up to do, even as it ships Koa on Nvidia's Nemotron and deepens its Claudeforce tie-up with Anthropic.

“Them slowing down a little bit is kind of good news because we have a lot of catching up to do.”