Understanding what drives progress in artificial intelligence (AI) is crucial for researchers, policymakers, and businesses. While advancements can come from better algorithms, more data, or increased computing power, the relative importance of these factors often remains unclear. This makes it difficult to allocate resources effectively—whether it's funding research, building infrastructure, or collecting data. A systematic analysis of how these components contribute to AI progress across different domains could provide valuable insights.

Breaking Down AI Progress

One way to approach this would be to select high-impact AI milestones—like AlphaGo for game-playing or AlphaFold for protein folding—and analyze how much of their success came from algorithms, data, or compute. For example:

• Algorithms: Did a new training technique or architecture lead to a breakthrough?
• Data: Was the improvement driven by larger or higher-quality datasets?
• Compute: Did access to more powerful hardware make the difference?

By comparing these factors across domains, patterns might emerge—such as whether compute is more critical for scientific tasks than games. Over time, trends could also reveal whether certain factors (like data) become more or less important as AI evolves.

Turning Insights into Action

The findings could help stakeholders make better decisions. Researchers might focus on algorithmic efficiency if data is already abundant, while policymakers could prioritize funding for compute infrastructure if it consistently drives progress. To test the idea, a pilot study could analyze a few well-documented domains (like chess and image recognition) using publicly available research. If successful, the framework could expand to other areas, with results shared through papers, datasets, or interactive tools.

This kind of analysis could fill a gap in existing AI research, which often focuses on individual breakthroughs rather than comparing progress drivers across fields. By providing a clearer picture of what fuels AI advancements, it might help guide future innovation more effectively.