Discussion on LLM Copyright and Product Direction
The discussions cover the positioning of AI products and the copyright issues surrounding LLM training data.
Key insights
User Needs-Oriented Critique: Products Should Focus on Usability
The experience summary points out that many AI product directions are misguided, emphasizing that users care about how to use the product to solve problems rather than the team's background, investors, or boasting. It suggests that 'teaching to fish' is better than merely showcasing capabilities, requiring products to directly provide tools or tutorials for problem-solving, and stressing that product design should be centered around actual user needs.
Fundamental Controversy Over LLM Training Data and Copyright
The discussion points out that almost all LLMs carry a risk of infringement according to traditional definitions, as their training data includes a vast number of books and sources (mentioning libgen), and questions whether all these datasets were acquired through purchase. It calls for LLM companies like OpenAI to unify solutions with content producers regarding copyright issues to clarify responsibilities and compensation methods.