微软首席执行官萨蒂亚·纳德拉在X平台发文阐述了他对高级AI模型安全风险的看法。1纳德拉主张应当假设所有AI模型都已被"破坏"，并呼吁业界不再接受将AI系统视作"嵌套黑盒"的现状。1
他提出了一系列具体的治理措施来应对这些风险。1这些建议包括及时披露安全事件、进行独立审计、确保数据可验证性以及实现模型隔离等措施。1纳德拉强调，需要建立更加透明的系统框架，并留下"防篡改的人类可读证据"，以确保AI模型的运行过程可被有效监控和验证。1
Microsoft CEO Satya Nadella has articulated a cautionary stance on advanced artificial intelligence systems, arguing that organizations must operate under the assumption that all AI models have been compromised. 1 In his statement, Nadella rejected the current practice of treating AI as opaque systems, calling instead for a fundamental shift toward greater transparency and accountability in model development and deployment. 1
To address these security and integrity concerns, Nadella proposed a series of measures designed to strengthen AI governance. 1 His recommendations include timely disclosure of incidents, independent audits, verifiable data practices, and the isolation of models from potential threats. 1 Additionally, he emphasized the importance of maintaining tamper-proof, human-readable evidence to ensure the integrity and traceability of AI systems. 1 This comprehensive approach reflects Nadella's view that the industry can no longer accept AI systems functioning as "nested black boxes" without visibility or accountability mechanisms. 1
评论
还没有评论，欢迎留下第一条。