liveBook · Manning

Data harvesting

As explained in How Large Language Models Work

Data harvesting is the large-scale automated collection of information from digital sources for subsequent analysis or model development.

Data harvesting is the large-scale, automated collection of digital information for later analysis or model development. In LLM development, it means gathering enormous collections of human-created content, often from websites whose users and publishers may not expect that use. section 9.1.2

The practice creates practical and ethical tensions:

  • It can provide valuable training material while contributing little to the advertising revenue that supports hosting websites. section 9.1.2
  • It may weaken the supply of future human-created content if creators lose income or motivation when their work is collected without acceptable compensation. section 9.1.2
  • It can provoke resistance even when licensing agreements exist, because users may object to the commercialization of their contributions. section 9.1.2

Stack Exchange and Stack Overflow

Stack Exchange hosts user questions, answers, and ratings across communities such as programming, system administration, mathematics, and tabletop gaming. After Stack Exchange sought payment from LLM creators, some Stack Overflow users deleted helpful answers in protest over their free labor being sold. section 9.1.2

How How Large Language Models Work covers it

How deep it goes
Defines the practice and traces economic and motivational consequences, but does not explain collection systems or implementation details.
Examples
Uses Stack Exchange and Stack Overflow, including deleted answers and proposed payments from LLM creators.
What you can do after
Evaluate how training-data collection can support LLM development while undermining platform finances, consent, and future content supply.
How it connects
Connects data harvesting with LLM training, user-generated content, advertising revenue, content licensing, and creator incentives.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of How Large Language Models Work. The section links point to the parts of the book it draws on.