AI offers some novel crystal materials that could form future chips, batteries, more Google DeepMind says it has developed an AI model capable of predicting millions of inorganic crystal structures that could potentially be used to form next-gen microprocessors, electric batteries, solar panels, and the like. Crystalline structures are made up of atoms arranged in a repeating pattern. They often exhibit …
I can totally believe that this is the way to go forward in creating new and useful materials to boost our technologies in every domain. Of course, as noted, a bit of fine-tuning is still required, but this can potentially bring many more returns than some stupid filter for smartphone pics.
And hey, if a machine finds us the proper compound for room-temperature superconductors, I won't complain.
We learn as much from mistakes and successes.
In my previous life, I worked in a fab. Awestruck by the smart people and wiz bang equipment, I was blown away by the simple process to make chips.
You start with a wafer. Overlay it with multiple layers of stuff, shoot it a few times light, plasma and chemicals and you have a chip. Test it. If it works then make it smaller with more chips on the wafer.
Test it.
If a percentage fails, change the process until it gets better and make it smaller.
Test it.
If a percentage fails, change the process until it gets better and make it smaller.
Test it.
If a percentage fails, change the process until it gets better and make it smaller.
And so on till it is as small as you want.
The key was shorten the time between build and test.
Is this not what the AI is doing, but virtually?
What immediately popped into my mind was that creating materials to order possibly isn't always as easy as it might seem. Example: Dolomite -- Calcium Magnesium Carbonate (CaMg(CO3)2). Simple enough molecule. One part Calcium, one part Magnesium, two parts Carbonate. Mix (cautiously) and (if necessary) cook. There is a LOT of dolomite on Earth. It is created in nature by replacement of one Calcium atom in the Calcium Carbonate in calcareous sediments by a Magnesium atom. Happens a lot. There are whole mountains made of the stuff.
Turns out to be very difficult to make in a lab.
Here's a reference https://pubs.acs.org/doi/10.1021/acsomega.1c04624
Doesn't mean that this AI effort or its trial and error old-fashioned equivalents are useless. Just means that things are likely to be (as is very often the case) more difficult than they seem.
I'm not a materials scientist, but it does sound a bit like an over-enthusiastic comp-sci major has bounded up to Paul Hollywood and told him they've generated 380,000 recipes for cake by working out every possible combination of eggs, flour and water...
How useful is this? You've made the finest green(*), it's sitting there sparkling on the lab bench... now what? How do you figure out if it is remotely useful, and under what circumstances? A lot of our currently exciting materials are only useful when you apply specific materials to their surfaces, after you've sliced and diced them in a particular way, then kept them at a specific temperature whilst some combination of electricity, light, radiation, water, gasses, other chemicals are applied in a very precise manner.
Worse still, you've got 379,999 other crystals lined up to test..
(*) Oh go on, you've got to make the joke.
They note that the outputs are then fed through traditional computer modelling software to verify them. It sounds like this is how they winnowed down 2.2 million concepts to a few hundred potential recipes, and further down to 51 sample test candidates.
Presumably part of this process involves the materials scientists having some notion of what's physically possible and also what may be materially useful.
We’re all for using computers to gain advantage and insight and to offer some guidance and comparison on the suitability of different options, especially when there are limitless possibilities. DeepMind, known internally as WheatAndChaff, does what a computer does, and is no doubt using more or less the same techniques as it does for the protein folding lucky bag. Can’t wait for DeepMind to offer suggestions for press releases regarding the use of computers to do some statistics work.
— Shaza DuPont —
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
