High-throughput experimentation

While many people jump at the idea of a Self Driving Laboratory, where AI for material discovery has a predominant role (at least in the way people talk), it is often forgotten that there's no path that does not go through high throughput experimentation as an enabler.

Even without any optimization algorithm, we should be able to run experiments continuously and in parallel. Generate reliable data that can be consulted later, shared, and expanded.

It is not only a problem of having the algorithms in place, but an acknowledgement that reality is messy.

Not two experiments will yield the exact same result, and dealing with noisy inputs and inconsistent results does not seam to be an easy task for deterministic algorithms.

The challenge is that not a lot of people are used to working in control systems, and even fewer have been exposed to scientific tools. While we have been training scientists to work without computers (I know, sounds like an exaggeration until you talk to a chemist), and engineers to work outside of science, that's the gap we must begin to close if we want to push for next generation tools.