Microsoft CEO Urges Companies to Reclaim AI Learning Rights: Stop Paying to Help Model Vendors Accumulate Experience

According to monitoring by Beating, Microsoft CEO Satya Nadella has once again published a lengthy article, urging companies to keep the learning outcomes generated by AI within their own hands. He referred to the current risk as a 'reverse information paradox': companies not only have to pay for AI but may also hand over their most valuable internal expertise. The more accurate a model aims to be, the more business information companies need to provide. Prompts written by employees, tools invoked, internal evaluations, and error correction records all contribute to unique company knowledge. However, under the current model, this information may also end up helping model vendors improve their products. Nadella criticized some model companies for believing they can learn from publicly available content on the internet while restricting customers from using model outputs to train their own systems, and potentially continuing to learn from customer usage records. If knowledge flows only to the supplier side, the manpower and expertise invested by companies could ultimately become assets of the model giants. He called on companies to take control of their evaluations, memory, operational trajectories, and fine-tuning weights, and to separate the agent orchestration layer from individual models. Even if a model vendor exits the market or raises prices, the business capabilities accumulated by companies will not disappear. A month ago, Nadella had already proposed the 'Human Capital and Token Capital' framework, warning companies not to outsource their learning capabilities. [Original Link]