How to Train a Machine Using a Knowledge Graph

The idea is exactly what Peter said before. We give a goal to a system, to an algorithm and say, build me a machine that does this,. And then you have, as I said, different steps. You have maybe the first step is designing a new component because you don't have a fitting component to build your machine. Then you use existing components, plus this component that you've designed to build a machine, and then you've programmed that machine. There we have huge potential.