Data Leverage and the Harry Potter test: Putting a value on data in AI models.

The value of data used for AI models and how to measure it is gaining attention. Researchers are exploring the concept of "data leverage" to determine the value of specific data pieces in large AI models. Recent papers by AI companies Anthropic and researchers from the University of Washington introduced methods to efficiently swap data and measure its contribution to model output. Understanding the value of data will enable communities to negotiate fair compensation for their contributions. However, legal concerns and copyright issues remain unresolved.

Posta un commento