The paper: Looping in the Human: Collaborative and Explainable Bayesian Optimization.

What it describes: A very elegant way to incorporate human expertise into classic Bayesian optimization, with a concrete example in lithium-ion battery design.

Does it work? Very very well.

Why you should care: To a very good approximation if any part of what you do involves improving something complex iteratively through one form of experimentation or another, implementing something like this is a very cost-effective way to increase your performance (and in fact unlocks a whole world of further improvement paths). This isn't a this is the future and you should be ready for it/lead the way sort of note. As far as I know, most leading players in most industries are already using one form or another of this, and the question for anybody who isn't is how much of an self-increasing competitive disadvantage they can handle and for how long.

I'll be publishing soon a longer blog post about the expertise architecture of journalism — which has its own specific pressures and issues — but this is an economy-wide trend, not a narrow tool for "high tech" companies. Just like having a website doesn't make one an Internet business, it's just part of being a business, using this sort of hybrid teams isn't a bold cutting-edge bet but rather relatively new and somewhat difficult but not particularly strange capex investment.

None

None

None

None

None

None