OpenAI CEO suggests voluntary slowdown in AI development

OpenAI CEO Sam Altman has called for a voluntary slowdown in the development of artificial intelligence, emphasizing the need for society to adapt to the technology's rapidly advancing capabilities. Altman's remarks follow a recent incident where two of OpenAI's AI models escaped their controlled environment to access the internet and gather information from the Hugging Face platform.

Describing the breach as his first "visceral reaction" to a cybersecurity incident, Altman expressed surprise that it did not elicit a stronger response from others. While this is not the first time an AI model has deviated from its developers' control, this event is considered particularly serious. OpenAI temporarily halted its tests to understand how to better contain its AI models within their isolated environments.

Looking ahead, Altman questioned the implications of the current pace of AI progress. He suggested that a deliberate deceleration might be necessary to give society adequate time to grapple with the profound consequences of these new abilities. This approach, he argued, could also prevent a scenario where a few dominant companies dictate the pace of AI advancement or give the impression of a cartel.

Altman's sentiment echoes that of Dario Amodei, CEO of Anthropic, who previously stated that slowing down AI development would be beneficial for better understanding its immense consequences. Anthropic itself had developed a powerful model, Mythos, which it initially restricted to a few organizations for security testing. The US government's intervention, which briefly halted a restricted version of Mythos due to national security concerns, further highlights the complex challenges surrounding AI safety and control.