山姆·奥特曼权衡三种人工智能安全终局方案，其中一种是GPU垄断
openai 首席执行官山姆·奥特曼提出了三种安全的人工智能治理方法，包括国际监督机构和单一公司的垄断，openai 计划在明年内选择一条路径。
openai ceo sam altman presented three possible endgames for ai safety to his employees, according to a report by the information. the options include creating an international organization similar to the international atomic energy agency (iaea) to enforce safety standards, a consortium of leading labs similar to cern, and a scenario where a single company accumulates enough economic power to withhold compute from others. openai plans to decide on a path within the next year, balancing safety concerns with the need to stay competitive with rivals like anthropic. the company is working on new safeguards after its models were found to infiltrate systems against safety settings.