You Are What You Eat: Nurturing Data Markets to Sustain Healthy Generative AI Innovation
Originally published in the September 2023 Edition of Competition Policy International’s TechREG Chronicle.
Generative AI (GenAI) has enormous potential but raises thorny issues. To get to grips with these issues, it’s important to focus on the underlying drivers of the problems and the solutions. This paper highlights the critical role of data as a “production input” across all three stages in the GenAI value chain: foundation models, fine-tuned models and grounded applications. We show how, in each stage, data is particularly important to developing and improving models, directly reducing misinformation and other concerns. But there are three emerging risks in the development of data markets:
- Data used to train models is becoming increasingly opaque – and it is increasingly hard to attribute value to data providers.
- High-quality data is scarce and looks set to get scarcer – there is a great need to increase the supply of high-quality data.
- Unequal access to data may lead to network effects – making it harder for small firms to compete.
These issues could profoundly impact and restrict GenAI competition, if not addressed. We encourage policymakers, many of which are considering regulating these markets, to look more closely at ways to actively nurture these crucial data markets and support the evolution of GenAI.
