Restructuring science for AI
Science institutions and training models must adapt as frontier models and AI agents reshape scientific work.
Why this is a top priority
The scientific process is continuously changing, now at an accelerated pace with AI, and institutions must continue to adapt to support new approaches. Though science was once primarily the purview of small, single investigator-led labs, it increasingly came to rely on larger, interdisciplinary teams and is now changing once again as frontier models and AI agents disrupt different parts of the scientific workflow. Though institutions are much slower to adapt, the federal government and philanthropies can play an important role in supporting new institutional structures that are designed around modern ways of doing science. These structures may emerge by modifying existing institutions or by building new ones.
Science institutions, like many others, must also confront a new reality where many tasks that were historically handed to trainees are now handed to AI agents. Training is still essential for producing the future experts that will direct scientific research. The federal government and philanthropies should continue to support trainees directly while also piloting new programs that teach trainees the key skills they will need to be successful in research.
What should be done
Federal science agencies
- Federal science agencies should pilot new AI-forward research approaches: For example, the NSF TIP directorate has launched X-Labs to support new platform technologies. X-Labs awards offer flexibility that allows AI methods to be incorporated from the very beginning of a research program. See full proposal.
Philanthropists and federal funders
- Philanthropists and federal funders should ensure that scientists can access the data and compute needed to perform cutting-edge, AI-accelerated research: NSF’s National Artificial Intelligence Research Resource (NAIRR) and DOE’s Genesis Mission provide some researchers with the compute they need to train new AI models, but demand is growing. What new programs are needed to ensure that scientists can access the compute and large-scale datasets needed to train models in areas neglected by industry?
- Philanthropists and federal funders should pilot new AI-forward training opportunities: AI agents can now do the work of graduate students in many scientific fields. How should training opportunities evolve to address this new reality?
- Philanthropists and federal funders should pilot new AI-forward research approaches: Pilot new institutional models designed around modern, AI-enabled scientific workflows.
Get in touch
If you are a grantmaker, entrepreneur, or policymaker interested in accelerating progress in restructuring science for AI, get in touch by emailing science-structures.tls@ifp.org. We are not able to reply to all requests, but we read them all.
You can join our talent database to work on restructuring science for AI here. We will use this database if we ever headhunt for talent to found or be an early employee in organizations working on this issue.
Join our Launch Sequence mailing list here to receive new proposals, important updates, and occasional opportunities straight to your inbox.
