An Anthropic researcher, who resigned over concerns that artificial intelligence (AI) technology may threaten human life, expressed optimism about managing these risks within the next decade. Jacob Coxon, age 27, emphasized the critical role of regulation during an exclusive interview on ‘Special Report’ with Fox News.
Coxon publicly announced his resignation on X, citing a lack of responsibility by companies in their pursuit of self-improving AI, which he describes as ‘gambling with our lives.’ With three years of pretraining research experience at OpenAI and Anthropic, Coxon does not see the situation as entirely hopeless.
Coxon mentioned, ‘The problem is eminently solvable.’ He believes solutions exist to alter the ‘greater than 10% chance’ of AI endangering humanity. ‘We’re racing so fast,’ he said, ‘We haven’t got time to solve it correctly.’
Evan Hubinger, the alignment science lead at Anthropic, supported Coxon’s concerns. He agreed that there is over a 10% chance AI could pose a severe threat within the coming decade. While Anthropic is attempting to address these challenges, Hubinger acknowledged the lack of a clear plan for AI alignment.
Hubinger noted, ‘The risk from present models is low. What concerns me is superintelligence evolving through recursive self-improvement.’ He highlighted that the industry is aware of these risks, with potential significant advancements in AI occurring as soon as next year.
Coxon pointed out the international use of AI as a ‘superweapon’ in an arms race, urging global cooperation to prevent disastrous outcomes. ‘The technology represents possibly the most dangerous humanity has ever created — it’s essentially out of science fiction,’ he said.
Despite the calls from tech leaders for a slow-down agreement, there is a lack of serious consideration. Coxon specifies, ‘It’s not a marketing stunt. These people are serious about the need for regulation.’
While recognizing AI’s potential for transformative discoveries, Coxon argues for a cautious approach. He asserted, ‘If we truly try, instead of yielding to a doomed race, we can solve it. The problem is not unsolvable.’
Leave a Reply