News
Microsoft CEO Nadella criticizes the "double standards" of AI model companies: while advocating "fair use" of data, while restricting others' distillation
2 min read
Source: ithome.com
IT House reported on July 13 that Microsoft CEO Satya Nadella recently posted a post on the X platform, implicitly criticizing the practice of training models by large AI model companies including Anthropic. Nadella said that some model manufacturers advocate the "fair use" right to use public data to train AI models, while objecting to other companies learning their model capabilities through model distillation (Distillation). This approach is obviously contradictory. The so-called model distillation refers to using a larger model with stronger performance to output results and training a new model with fewer parameters and weaker capabilities. Nadella wrote: "It is certainly important to innovate that model providers have fair use rights to train models based on public data. But I think it is ironic to enjoy this right, then impose strict restrictions on model distillation, and reserve the right to continue learning using customer usage and interaction data." He further said that if the learning of knowledge can only flow in one direction, then the people who ultimately make all the money will be the ones who master the AI infrastructure, and the people who actually create the knowledge will not benefit from it. Currently, leading AI companies including Anthropic, OpenAI and Google DeepMind rely on a large amount of content created by others to train their models. The capabilities of models such as ChatGPT, Claude, and Gemini are based on publicly available text, images, and other Internet data. In recent years, a large number of companies and individuals have filed lawsuits against these AI companies for training AI by scraping content without authorization. Although Nadella did not name him, it is generally believed that his remarks were mainly directed at Anthropic. Earlier this year, Anthropic CEO Dario Amodei publicly accused Chinese AI companies of “stealing”