344
Productivity & Workflow355
Automation & Workflow224
Software Development251
Marketing & Growth192
AI Infrastructure & MLOps174
Writing & Content Creation203
Data & Analytics141
Design & Creative170
Photography & Imaging156
Customer Support131
Sales & Outreach125
Voice & Speech135
Education & Learning131
Operations & Admin87
Microsoft CEO Satya Nadella says companies may be “paying twice” for AI by paying fees and also giving away valuable business know-how through prompts and feedback.
In short: Microsoft CEO Satya Nadella says companies using paid, closed AI models may be giving away valuable business knowledge when they use them.
In a blog post published Monday, Satya Nadella warned companies about a hidden cost of using proprietary AI models, meaning AI systems you pay for but cannot fully inspect or control. He argued that businesses pay money for AI usage, and also “pay” with information they must share to get good results.
Nadella said the more a company wants an AI model to help, the more it has to feed the model sensitive context about how its business works. That includes prompts (the instructions people type), the tools connected to the AI, and especially the corrections employees make when the AI is wrong. He called this “exhaust,” like the trail of clues left behind when people use a system (similar to leaving notes in the margins of a manual).
He also criticized limits some AI providers put on “distillation.” Distillation is when you use one AI model’s answers to help train a smaller model, like learning to cook by tasting a chef’s dishes and recreating them at home. Nadella argued it is unfair for AI companies to train on public internet data, but then block others from learning from their models.
For everyday people, this affects prices, privacy, and competition. If businesses feel they must keep switching away from certain AI tools to protect their know-how, they may spend more on tech and pass costs along. Nadella’s comments also point to a future where companies try harder to keep AI work inside their own systems, rather than sending questions and internal details to outside providers.
Source: TechCrunch AI