Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

>>> Combining a series of new microscopy techniques, the group present a complete picture of the nanoscale chemical, structural and optoelectronic landscape of these materials

It's not even relevant to call it "microscopy" anymore, we require a new term. It's a complete thin film atlas of all interacting forces of nature. Better data for the models, means higher fidelity simulations.

The question is can AI predict new materials? Can a simulation be sophisticated enough to predict say, high temperature superconductivity in rare earth cuprate perovskites?



I'm not sure if this is exactly what you're talking about, but I read something a while back about using AI to predict if certain metallic glass alloys will have useful properties: https://phys.org/news/2018-04-artificial-intelligence-discov...


If we are going to speculate, then the question is, why couldn't AI do it? What kind of fundamental limitation would it hit? Data? But we could get a ton of data from already existing software[^1]. It is slow, but I have the feeling it wouldn't require as much computing as GPT-3, and it would perhaps be enough to train more efficient neural networks that can do the actual search.

Because how important is for human life, a compiler industry that finds ways to translate complicated simulations to AI algorithms could be the next big thing.

[^1]: https://en.wikipedia.org/wiki/List_of_quantum_chemistry_and_...




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: