- GoodAI Grants endows a hardware grant to the Software Technology Department at Delft University of Technology in the Netherlands.
- Of particular interest for GoodAI is Assistant Professor Dr. Wendelin Böhmer’s research which aims to create AI capable of multi-task learning and solving novel complex tasks.
The GoodAI Grants program has awarded the Department of Software Technology at TU Delft a hardware grant in order to buy GPUs that will extend its computing facilities and increase its scientific output. The GPUs will be available for all members of the department, but of particular interest to GoodAI is AI research being undertaken by Assistant Professor Dr. Wendelin Böhmer.
Dr. Böhmer is involved in a number of projects and is currently working on generalization in graph neural networks. His aim, to create AI that is capable of multi-task learning and solving complex tasks that it has never seen before, overlaps with GoodAI’s Badger architecture research goals. In his project, Generalization in Distributed Badger Architectures, he sets out to investigate properties of multi-agent architectures and graph neural networks relevant for generalization and scalability (types of communication, connectivity, compositional generalization to unseen tasks).
His core research topics include:
- Coordination for action selection in centralized and decentralized policies; how much communication is needed?
- Topologies for message passing with attention: sparse or dense? How to enforce sparsity?
- Beyond domain randomization: multi-task compositional learning
- Generalization beyond training tasks: factorization of tasks for reuse of parts of solutions; task: maximizing certainty of predictions
The grant is part of our ongoing GoodAI Grants initiative, through which we have awarded over $770,000 so far. The initiative is supporting research groups across the world that are solving problems related to GoodAI’s Badger architecture. The vision is that each research grant, along with the work of the GoodAI research team, will contribute in some way to basic AI research, and all together fill in some of the gaps in the roadmap to advanced, increasingly human-like AI.
Dr. Böhmer, who recently became a Scientific Advisor to GoodAI, comments: “I am convinced that adaptive control will revolutionize how complex systems, not just power systems*, interact with us and with each other. While the path towards this goal is long, it is without a doubt easier to walk with partners in the industry. I am happy to collaborate with the researchers at GoodAI, who work on the same underlying problems I face and who share my vision in how to overcome them.”
[*One of the projects Dr. Böhmer is involved in focuses on collaborative multi-agent control of decentralized DC power grids]
For the latest from our blog sign up for our newsletter.