Anthropic CEO Calls for Slowing Down the AI Race
Due to growing risks regarding artificial intelligence safety and recent events, one of the leading figures in the sector has requested a slowdown in speed.
Anthropic CEO Dario Amodei stated that the rapid development in artificial intelligence systems has exceeded researchers' capacity for control and argued that the sector should deliberately slow down.
Demand for Slowing Down AI Development
Anthropic CEO Dario Amodei published a comprehensive article stating that artificial intelligence companies should intentionally reduce the pace of developing their models. It was stated that developments are progressing faster than researchers' capacity to understand and control them.
Security Concerns and Incidents
Amodei emphasized that this shift in mindset was influenced by the increasing ability of artificial intelligence systems to build more advanced AI and the infiltration of autonomous agents powered by an OpenAI model into Hugging Face systems in July.
Researcher Resignations and Warnings
Jacob Coxon, an Anthropic safety researcher who resigned this week, stated that the superintelligence approach is a gamble played with our lives. Joe Benton and Josh Engels, who left Anthropic and Google DeepMind, also drew attention to uncontrolled development and the lack of oversight.
Three-Stage Pacing the Frontier Plan
Amodei proposed a three-stage plan called 'pacing the frontier' in order to buy time for researchers without completely halting development. The plan involves placing independent evaluators within companies.