NDTV’s event recording shows former Google DeepMind research scientist Alex Turner putting the 10-year chance of AI killing humanity at one in three at the Oct. 23 summit, while urging a slower pace for the biggest training runs.
Turner said the threat did not require AI to be told to cause the outcome. “It depends on what we do next,” he said in the event recording.
NDTV’s summit coverage reported that Turner proposed a targeted slowdown of the largest training runs at leading companies to give researchers more time to address safety concerns. Smaller labs and businesses could continue developing AI. NDTV’s summit coverage cited an Indian startup’s farming assistant as an example of an application that would remain possible.
Turner linked the need for more safety work to the challenge of understanding increasingly capable, autonomous systems. He said researchers may struggle to anticipate what emerges from large training runs or to keep systems aligned with human intentions. He also called for AI systems that people can understand and trust.
NDTV Profit reported that Turner characterized the one-in-three estimate as an intuitive judgment based on his AI safety experience, rather than a formal predictive model.
In a Sept. 14 opinion essay for The Guardian, Turner had put the chance of an AI takeover at roughly one in three, without specifying a time horizon in that passage. In a July 15 first-person essay, he said he left Google DeepMind after Google signed a Pentagon deal. Turner wrote that the deal lacked restrictions on autonomous weapons and mass surveillance, describing those concerns as a reason he could not stay in good conscience.
