Areas like computer vision, it would be valuable for governments to be able to not just monitor progress in underlying techniques, but to understand how the safety and robustness of these systems are evolving relative to the underlying techniques, and which actors are contributing to progress in different areas. Policymakers could build on existing datasets such as the ImageNet dataset to evaluate and monitor which research groups are capable of further developing the state-of-the-art, helping to make governments aware of useful clusters of expertise.
