Does AI change how much my data is worth?

🌈 Abstract

The article discusses the potential value of individual personal data in the context of the growing market for training data for large language models (LLMs). It explores the historical view of individual data as commercially insignificant, the emerging market for buying and selling consumer data for LLM training, and the challenges in pricing this data. The article also considers the possibility of individuals directly benefiting from the value of their data through data collectives and cooperative models.

🙋 Q&A

[01] The Value of Individual Personal Data

1. What is the traditional view of the commercial value of individual personal data?

• Individual datasets have traditionally been viewed as commercially insignificant because the market only attributes value to data in the aggregate.
• Whole empires have been built off of our collective digital exhaust, but this value has never really trickled down to people like the author.

2. How is the value of individual data changing with the rise of LLMs?

• As more LLMs announce partnerships with media companies in exchange for training data, the author wonders whether the value of their personal dataset is going up.
• The author notes that a separate (though related) market is emerging to facilitate the buying and selling of consumer data for the purpose of training LLMs.

3. What are the challenges in pricing individual personal data for LLM training?

• Mira Murati of OpenAI acknowledges the difficulty in determining the value of a specific amount of data and how much value it creates in a trained model.
• The author notes that it feels "a bit like shooting from the hip" when it comes to the pricing numbers discussed, such as Photobucket's proposed rates.

[02] Potential for Individual Benefit

1. What are the author's thoughts on the likelihood of platforms compensating individual users for their data?

• The author believes it is unlikely that platforms will compensate individual users in the form of cents and dollars, as it would be inefficient and too likely to end up as "selling your house for firewood."

2. What alternative models does the author see as more promising for individuals to benefit from the value of their data?

• The author is paying attention to less obvious business models that enable people to band together and form data collectives in which they own a stake, such as Vana and Hive Mapper.
• The author believes that data collectives and cooperative models that ensure individuals who contribute their data can directly participate in the benefits of its use are more likely to enable individuals to benefit from the value of their data.