Nadella is suspected of quarreling with the big model head player from a distance: he is only allowed to crawl the data himself, and others are not allowed to distill the model

Zhidongzhi Compiler | Eggplant Editor | Cheng Qian Zhidongzhi reported on July 13 that Microsoft CEO Satya Nadella posted a long article on X yesterday titled "The Reverse Information Paradox", questioning the "learning asymmetry" phenomenon that currently exists in the AI ​​industry. This article by Nadella was considered by the foreign media Business Insider to be a satire on the "double standards" of large model manufacturers such as Anthropic and OpenAI in the AI ​​training mechanism. On the one hand, these large model manufacturers allow themselves to use public data and user interaction data to continue learning, but on the other hand, they restrict other companies from obtaining model capabilities through model distillation. Previously, Anthropic has repeatedly publicly condemned other AI companies for copying Claude’s capabilities through model distillation, and has repeatedly called for strengthening the protection of model capabilities. Nadella pointed out in the article that under this mechanism, in order to use AI capabilities, enterprises need to continuously provide their own data, processes, feedback and professional knowledge to the model, and these contents may also become an important source of continuous learning for the model. If the learning process only flows one-way to the model provider for a long time, the knowledge value created by the enterprise may eventually continue to be concentrated in the party that masters the AI ​​infrastructure. Based on this, Nadella proposed that enterprises in the AI ​​era need to establish a new "Trust Boundary" to ensure that data, feedback, evaluation systems and organizational knowledge can be deposited within the enterprise to form their own continuous learning and intelligence accumulation capabilities. The following is a compilation of Nadella's full text on "The Reverse Information Paradox": In the era of intelligence, how should companies protect their core intellectual property? Nobel Prize winner in economics Kenneth Arrow once proposed a paradox in the information market. He once pointed out: “For buyers, the value of information is unknown before obtaining the information; but once the information is obtained, the value of the information is unknown to the buyer.