新一代隐私计算技术来了？蚂蚁密算：让系统运维者也没法窃密

Core Viewpoint - The increasing value of data emphasizes the importance of its circulation, with a focus on ensuring data security and preventing misuse when it leaves the control of its providers [1][5]. Group 1: Technology Overview - Ant Group's Ant Misan has been awarded the "Outstanding Contribution Award" at the World Internet Conference for its potential to address data security concerns [1]. - Ant Misan utilizes a new generation of privacy-preserving computing technology called "secure computing," which ensures that data is "available but invisible" during processing, thus preventing data leakage and misuse [1][4]. - The technology allows for effective control over data even when it is shared across different entities, industries, and regions, overcoming limitations of traditional plaintext computing [3][5]. Group 2: Market Potential and Future Applications - The cost of secure computing has decreased to below 1.5 times that of plaintext computing, with secure model inference costs dropping to under 1.2 times traditional inference costs [6]. - High-value and sensitive industries such as finance and healthcare are expected to be the first adopters of secure computing, leading to broader applications across various sectors as infrastructure develops [6]. - The establishment of a research group within the National Data Standardization Committee aims to promote the standardization of secure computing, marking this year as a potential turning point for market-scale applications [6]. Group 3: Addressing Challenges in AI Models - The phenomenon of "hallucination" in large models is linked to the compression of human knowledge, which can lead to inaccuracies when the model is faced with complex queries [7][8]. - Effective problem decomposition and the integration of programming languages with natural language are proposed as solutions to enhance the reliability of large models, potentially increasing efficiency in the data field significantly [8].