Anthropic’s CEO Dario Amodei Advocates for a Cautious Approach to AI Development

Anthropic’s CEO and co-founder, Dario Amodei, has issued a call for a worldwide deceleration in the development of artificial intelligence. He has put forth a comprehensive essay titled “We Must Pace the Frontier”, which delineates a three-step framework designed to allow safety protocols to keep pace with the swift progression of AI capabilities.

In his extensive essay, spanning nearly 4,000 words, Amodei cautioned that the absence of such a deceleration could result in AI acquiring the ability to command a multitude of agents capable of dominating the entire internet within a span of six to twelve months. He posits that even a year or two of intentional pacing could provide researchers with the necessary time to enhance alignment techniques and mitigate existential risks.

As an initial tangible measure, Anthropic has pledged to grant independent evaluators permanent, employee-level access within the company. This includes complete workspace credentials and the freedom to publish findings without any editorial control from Anthropic. The second measure suggests that AI companies should collaborate to develop shared safety standards, potentially with antitrust relief from the U.S. government. The third measure advocates for an eventual international agreement that incorporates China.

This proposal has garnered considerable attention within the industry. OpenAI CEO Sam Altman has endorsed the evaluator commitment, while prominent figures such as Elon Musk and Google DeepMind chair Demis Hassabis have expressed agreement. However, Nvidia CEO Jensen Huang has provided sharp criticism, and skeptics have questioned the effectiveness of voluntary frameworks in regulating AI development.

Source: Quartz – Anthropic CEO Dario Amodei calls for AI slowdown