英国AI部长Kanishka Narayan将宣布加强英国网络防御的举措，同时呼吁开发前沿AI的公司加强安全防护1。这是Narayan首次以AI内阁级别官员身份就该议题发表演讲1。此前，英国情报机构MI5、MI6和GCHQ已向内阁通报了AI对国家安全构成的威胁1。
关于AI安全风险的评估各异。有专家警告，敌对国家可能盗取未发布的尖端AI模型，并指出OpenAI的安全防御不足以防止此类盗窃1。另有观点认为存在"50%的可能性"人类被超级智能AI消灭，并建议立即停止先进系统的开发1。微软首席执行官Satya Nadella则呼吁AI模型设计配备"应急制动"机制，将超级智能视为内部风险1。
在国际层面，英国首相表示将利用明年G20主席国地位推动建立"单一套全球原则和标准"的AI安全协议1。然而，英国目前尚未加入由芬兰、法国、德国、澳大利亚、新加坡、阿联酋和爱尔兰领导的国际AI安全标准倡议1。
Britain's newly appointed AI cabinet minister, Kanishka Narayan, will announce measures to reinforce the country's cyber defences while calling on developers of cutting-edge artificial intelligence to intensify their safety protocols 1. The move follows warnings from UK intelligence agencies about the national security threats posed by AI, with MI5, MI6, and GCHQ recently briefing the cabinet on these risks 1.
Geoffrey Irving has warned that hostile nations could steal unreleased advanced AI models, arguing that OpenAI's existing safety protections are insufficient to prevent such theft 1. Irving has also predicted that humanity faces a 50% probability of being eliminated by superintelligent AI and has recommended an immediate halt to the development of advanced systems 1. He further predicted that "on the current trajectory, we might achieve superintelligence within 12 months" 1.
The UK government has signalled its commitment to advancing global AI safety standards. The Prime Minister has indicated plans to leverage Britain's G20 presidency next year to promote a unified set of international AI safety principles and standards 1. However, the UK has not yet joined an international AI safety standards initiative led by Finland, France, Germany, Australia, Singapore, the United Arab Emirates, and Ireland 1. Additionally, Microsoft CEO Satya Nadella has called for AI model designs to incorporate "emergency brakes," treating superintelligence as an internal risk management concern 1.
评论
还没有评论，欢迎留下第一条。