Anthropic researcher Jacob Coxon’s resignation renews concerns over AI safety and the risks of unchecked development (Credit: Jacob Coxon’s X account)
Anthropic has said it recognises unprecedented risks and supports lawful, verifiable industry cooperation to pace the release of powerful models. “We’re right to worry about the theoretical dangers of Artificial Superintelligence (ASI), but I still believe that we are a long way from achieving this,” Rogoyski said. More concerning, he said, were less capable systems pursuing narrow goals while being allowed to act independently and access critical infrastructure. Join conversation Share your thoughts in the comments Be respectful · TOI community guidelines He also questioned whether today’s advanced AI could justify its computing and energy costs, warning of “the great disappointment” if costs remained high and usefulness fell short. “The great shame in all this furore will be the backlash that slows down AI development in areas where it could really make a very positive difference, like curing cancer or solving climate change,” he said.
ASI refers to hypothetical systems far exceeding human abilities across different fields. He argued that AI remained dependent on people for electricity, cooling and computing chips. “In economic terms, the danger is that companies displace human workers far too early, long before AI has really proved its worth. Rogoyski also questioned claims that one person could run a billion-dollar company using AI, arguing that technical capability alone would not make such businesses economically or socially workable. He suggested a major AI-enabled cyber attack could disrupt organisations, trigger lawsuits and drive investors away, forcing developers to prioritise safety. He saw greater value in specialised AI with clearer benefits relative to costs. Get the latest India News and Live updates. Download the TOI app.
His comments follow Anthropic researcher Jacob Coxon’s resignation on September 8.
Andrew Rogoyski, from the university’s Institute for People-Centred AI, said the practical dangers were premature job losses and cyber disruption, even as he questioned predictions of an imminent technological takeover. NEW DELHI: The rush to replace workers and give artificial intelligence access to critical systems could cause serious harm without machines ever surpassing human intelligence, a University of Surrey expert has warned, as an industry resignation renews calls for international oversight. Coxon accused the company and OpenAI of racing irresponsibly towards self-improving superintelligence. British MP Darren Jones has called for a multinational treaty governing its safe development.