I'm quite skeptical of the interdisciplinary approach. We already know that it's possible to create useful neural networks without any "preprogramming", i.e., creating AlphaGo zero without beginning with any knowledge of Go. An interdisciplinary approach seems backward in that it premises that we need to know something about the problem before we can build the tools to find the solution. That's increasingly not the case, and I think what will ultimately happen at this College is that CS researchers will wall themselves off from the rest, who will just slow them down. If you want to bring AI to other fields, educate the people in those fields or hire an ML engineer.