Jacob Coxon, a former researcher at OpenAI and Anthropic, testified Monday before the New York City Council that humanity will likely lose control of advanced AI, which could result in human extinction. Daniel Kokotajlo, another former OpenAI researcher subpoenaed to testify, stated that AI companies may not know when their safety work has failed and compared the field more to psychology than engineering. Alex Turner, a former Google DeepMind researcher, placed the chance of AI takeover at roughly one in three. The City Council is considering legislation that would require outside validation and the ability to manually shut down AI systems, with fines of $25,000 per violation. When asked if their companies carried insurance against catastrophic risks, no representative from the AI companies present raised their hand.
Source: Read the original article
