AI CEOs demand industry slow the pace of advancements after ex-Anthropic employee warned tech may kill us all

The heads of AI giants Anthropic and OpenAI are demanding other industry frontrunners join them to “slow the pace” of advancements before rogue swarms take over the internet.Anthropic CEO Dario Amodei published a chilling essay on Saturday urging AI trailblazers across the globe to commit to a three-part plan to prevent an AI catastrophe.

OpenAI CEO Sam Altman then echoed the urgent sentiment.“Given the accelerating rate of AI capability development, it’s my worry that in 6–12 months such a swarm could be capable of taking over the entire internet with a persistent botnet (potentially causing hundreds of billions of dollars in damage), and that the scale of damage would continue to increase from there if AI becomes more powerful without the necessary guardrails,” Amodei wrote.“We must slow the pace at which we improve the capabilities of AI models.Progress will still seem fast, and we must make wise use of the time we gain.”Amodei, 43, said the first step that should be universally adopted is granting outside evaluators employee-like access to companies to monitor their practices, ensuring that safety and ethics are prioritized.“Anthropic is unilaterally committing to this step now. We intend this to be part of a broader push to redouble efforts on our safety and alignment work,” Amodei said.

The next step is for AI companies leading the potentially dangerous charge to come together to set universal standards about pacing, he wrote.He then called for “global coordination” between the US and authoritarian governments like China, “to the extent possible.”“The measures I propose to advance the frontier at a safe pace will not be easy.But I believe we owe it to humanity to try,” Amodei concluded.

Altman reacted to the essay on X.“I agree with Dario that we need to pace the frontier.

This has been a primary topic of discussions we’ve had at OpenAI in recent weeks,” he said.“Committing to having independent evaluators with ...