Seán Ó hÉigeartaigh on Frontier Model Risks

Seán Ó hÉigeartaigh, an academic at the University of Cambridge who directs the AI:FAR programme, posted that absent regulation the risk from internal frontier models over the next 12 months will be set by which developer holds the most capable models and which is least aware of risks or employs practices such as risky RL environments. The post links to the Oxford Martin School paper on the unilateralist’s curse and a principle of conformity.