The alloy, called GRCop-42, is made of three metals – copper, chromium, and niobium.

I'm not sure how AI/ML would help identify working printing configurations. Seems like plain old simulations would have been more useful.

To clarify the definition of "AI" that they're referring to: they're using Bayesian modelling. There's not a single neural-network in sight in this paper.

It's actually a great demonstration that you don't need "AI" and a cluster of expensive GPUs to solve this kind of practical optimisation problem. We've had these kind of statistical modelling techniques for decades already.

These kinds of headlines are being used to justify the massive data-centre build out and investment, despite the fact that the amount of computation they needed to do for this paper was just normal number crunching that data scientists have been doing on their laptops since forever.

The paper: https://ojs.aaai.org/index.php/AAAI/article/view/41428/45389

This paper develops a Bayesian Experimental design for AM (aka BEAM) approach

BEAM is inspired by work on active search ( Jiang et al. 2018 )

The referenced paper (Jiang et al 2018) is the best place to read about the method that they used.