RDM & workflows for Artificial Intelligence
There will be no widespread adoption of AI for research until sound scientific methods have been established. Transparency and reproducibility are vital. SURF aims to identify and share the right workflows and tools for data management and AI.
What is the current challenge we are looking at?
Dealing with data: research communities miss out
Artificial Intelligence is a new instrument in the scientist’s toolbox. At the moment, there is a huge gap in AI proficiency between research fields, and between groups within those fields. Large fractions of those research communities miss out on valuable opportunities.
One of the fundamental issues is that AI projects involve more and more data, compared to traditional software development. This creates unique challenges in maintaining traceability between inputs and outputs, in particular when experiments with various models are involved. While researchers have long been trained to maintain proper lab notebooks to insure reproducible results, this is not yet the case when using machine learning.
What are we working on this project?
As a first step, SURF is analyzing what tools are available, both open source and proprietary software. By experimenting with them, we investigate how they may help scientists do research with AI, with minimal overhead. And what kinds of new approach they could enable.
Helping researchers to get started with AI
An important step will then be to share this knowledge to increase the uptake of AI. After analyzing the strengths and weaknesses of different approaches, we’ll share our insights with the different research communities to help users get started in the most effective way possible.
AI for all
We aim to make AI-based research available to many more scientists than are using it now. By establishing norms and helping researchers with their workflows, we can save them a lot of time and energy. They won’t have to make the same technical mistakes many people have made before them.
What are the main activities?
- Experimental research into available tools
- Introductory publication, to help users navigate through a synthetic use case
- Outreach activities, presenting applicable workflows in the SURF community.
Who are we collaborating with?
Based on the ongoing investigations and experiments, formal partnerships will be established at a later point.