Christiano Warns of Catastrophic AI Risks at OpenAI

Paul Christiano, a leading AI safety researcher, recently joined the OpenAI non-profit board and issued a stark warning regarding the future of the industry ^[1]. Christiano stated that current efforts are not sufficient to prevent a "catastrophic and irreversible loss of control" ^[1][3]. He estimates there is a 15% probability of a civilizational catastrophe caused by AI within the next three years ^[1][2].

The researcher highlighted specific difficulties in aligning recursive self-improving systems ^[1]. This warning comes as OpenAI continues to expand its board to address safety concerns ^[2]. Christiano's comments suggest that the risk of losing control over advanced AI models is a near-term challenge that requires urgent industry attention ^[3].