AI Researcher Quits Top Firm, Warns of Existential Threat by 2030
Jacob Coxon, a 27-year-old British researcher specializing in training advanced AI models, has resigned from the leading artificial intelligence company Anthropic, issuing a stark warning about the industry's trajectory. Coxon, who previously worked at OpenAI, expressed deep concern in a viral post on X (formerly Twitter) that both companies are "racing straight to self-improving superintelligence and gambling with our lives."
He claims that AI developers, despite projecting a calm public image, privately harbor anxieties about losing control of increasingly powerful AI systems. Coxon asserts that these future systems could surpass human capabilities, access any data, instigate radical societal changes, and rapidly accumulate resources, posing an existential threat to humanity potentially by the end of the current decade.
Coxon highlighted a perceived difference in awareness between OpenAI and Anthropic, suggesting that while OpenAI has not fully grasped the severity of the situation, Anthropic is acutely aware of the potential destructive consequences. However, he stated that Anthropic's management continues to push forward, believing they must win the race at all costs, a strategy Coxon labeled as an "arrogant and extremely dangerous gamble."
The researcher pointed to a contradiction within Anthropic's leadership, noting that CEO Dario Amodei and others have previously warned of AI dangers and called for slower development, yet the company's rapid advancement persists. Senior scientist Evan Hubinger reportedly acknowledged the significant possibility of AI destroying humanity and admitted the company lacks a clear plan for controlling such systems.
In related news, the article mentions Elon Musk's recent shift in stance regarding Anthropic, having previously called its models a threat but now praising the company as a quality global leader. It also notes Anthropic's past consideration of acquiring the Israeli startup Decart AI, which specializes in maximizing AI chip capabilities, though the deal did not materialize.
