AI researcher Jacob Coxton announced his departure from Anthropic on September 8 (US time). In a social media post, Coxton stated that as the pace of AI development accelerates, time to ensure the safety of AI systems is limited, and that the next 1, two years, will be a critical turning point for humanity.

In an interview with WIRED, Coxton expressed concerns that the movements of Anthropic and its competitors are at a stage that will determine the fate of humanity, and that the development race could sacrifice safety and research rigor. Furthermore, he pointed out the risk of AI being used for biological or cyberattacks and suggested that OpenAI and Anthropic must collaborate on measures such as restricting "recursive self-improvement," where AI develops new AI by itself.

Against the backdrop of accelerating AI capabilities, which are approaching levels that surpass humans in fields such as hacking, Coxton is calling for the need to adjust the pace of development and establish international coordination frameworks.

Source: