news.volyx.in

Who's afraid of Chinese models? (stratechery.com)

995 points by mfiguiere · 38 days ago · 910 comments on HN

Article summary

The article discusses the implications of Chinese models, such as Kimi K3, approaching the state-of-the-art in terms of capabilities, and how this affects the AI market. The author argues that the reaction to Chinese models is overblown from an economic perspective, as the cost of serving AI models is not zero and is related to the cost of goods sold (COGS). The article also explores the concept of intelligence as a commodity and how it will be affected by the increasing availability of AI models. The author suggests that the market will eventually treat intelligence as a commodity, with suppliers competing on cost structure.

Main themes

  • AI market dynamics
  • Commodity markets
  • Cost structure
  • Chinese models
  • Intellectual property
  • Regulation

What commenters say

  • The concept of distillation is not inherently bad and can be seen as a form of value extraction, but it raises questions about intellectual property and the rights of model creators.
  • The government can pass laws to regulate terms of service and make certain provisions unenforceable, which could impact the AI industry.
  • Some argue that distillation is a form of value add, as it requires selecting valuable parts from a large amount of data, while others see it as simply copying existing work.
  • The line between value add and value extract is blurry, and it's difficult to determine what constitutes original work in the context of AI models.
  • The use of Chinese models could become a value add if they are used to create new and useful applications, rather than simply copying existing ones.
  • Some commentators believe that non-compete clauses in software licenses are unfair and should be made illegal, as they stifle competition and innovation.
  • Others argue that such clauses are necessary to protect intellectual property and prevent the theft of trade secrets.
  • The concept of intelligence as a commodity raises questions about the role of regulation and the need for a framework to govern the AI industry.