A series of high-profile warnings about artificial superintelligence risks emerged over the past 48 hours, with prominent figures comparing potential dangers to nuclear weapons. Former UK Defence Secretary Des Browne and UC Berkeley computer scientist Stuart Russell presented concerns to British lawmakers, while a senior Anthropic researcher stated he believed there was greater than a 10% probability that AI could kill all humans within the next decade. The warnings came during parliamentary discussions convened by Control AI, an advocacy organization pushing for international oversight of the technology.
The concerns gained further momentum when Jacob Coxon, a 28-year-old researcher at Anthropic who previously worked at OpenAI, resigned claiming neither company was acting responsibly regarding AI safety risks. Coxon stated that the companies were “gambling with our lives” and operating in a competitive race to develop advanced systems. In response, Evan Hubinger, an alignment scientist at Anthropic, acknowledged concerns about the company’s preparedness should artificial superintelligence be achieved. The UK government confirmed that Anthropic declined to submit its latest model, Mythos 5.1, to the country’s AI Security Institute for pre-release testing.
Political responses escalated on both sides of the Atlantic. Labour MP Darren Jones wrote to Prime Minister Andy Burnham and international organizations calling for a multinational treaty governing superintelligence development with a safety-first approach. In the United States, Senator Bernie Sanders renewed calls for congressional regulation, citing polling showing 81% of Americans believed politicians should take action on AI safety.
Not all voices aligned with the most dire predictions. Dr. Andrew Rogoyski from the Surrey Institute for People-Centred AI suggested current systems remain far less versatile than humans and predicted potential disappointment with the technology’s practical utility. David Barber, director of a state-backed UK research lab, cautioned against overly restrictive approaches while acknowledging the need for better safety controls. The discussion reflected broader tensions between those advocating precautionary measures and those concerned about stifling beneficial AI development.
Article Attribution | Read More at Article Source
Article summary produced by Claude AI