gjmulhol

Apparently MIT didn't get the memo on atoms: "Molecules are among the smallest building blocks available for making next-generation devices."

Not only is this from 2018, but Yaghi himself demonstrated MOFs in 1995 so "new" is pretty relative.

I have very mixed feelings about this. It is an incredible amount of trust in Apple (both security and actual execution, esp given their history in AI), but it would be such a massive upgrade to individual user security that it is hard to comprehend.

This is really cool...these LLM-like models in technical disciplines are going to open up he new frontiers in science.

Come join our team -- super exciting technical problems of both the engineering/development and chemical engineering/materials science varieties!

This is a good list with some good advice.

As a personal anecdote, I had a very well-known investor tell me that my business would never make more than $5m per year (not venture scale). And this was even knowing that my business is a pretty standard enterprise software business model. We are well past that at this point, so I am proud to have proved him wrong.

This is cool in concept, but the live demo doesn't seem to be working. Or maybe I am using it wrong?

Cool, and definitely useful for complicated data structures

One of the big questions is whether we can extract these minerals from existing devices that are out of use/have been disposed of. Ideally, that means that we are designing new devices with disassembly in mind. Everyone from Apple to Huawei is talking about doing this, but the proof will be in the pudding.

Some of the comments are slightly inaccurate. We actually don't know where all the deposits are, and there was just a large discovery of rare earth elements off the coast of Japan. Much rare earth extraction has been happening in China because of relatively lax environmental standards that have reached back decades, so the US, Australia, and many other countries with deposits can't compete due to the costs of labor and complying with environmental standards and regulations. Mountain pass is one of a few deposits in the US, and happens to be one of the larger ones that is sitting idle. There are lots of mines globally and could be more, but economically they need to be attractive.

Copper is another interesting element. It is exceedingly hard to extract from existing devices because it is buried inside the chips, boards, etc. You can recover about 25% of the copper in a device, or about 3% of the total mineral content of a device when you extract copper. Widespread copper mining has destroyed parts of the Atacama desert in Chile and is a really nasty process.

Finally, work has been going on for a long time to replace rare earth elements or dramatically reduce how much is used. In some cases, you need small amounts because you need those f-orbital electrons but you can get away with using creative coatings instead of large volumes of materials. In other cases, you can replace them by multi-layering other materials to approximate their performance. There are huge opportunities here, but we are a couple decades away from seeing any real sea change in what our devices are made of.

Building physics and chemistry domain specific machine learning systems at Citrine Informatics. It's the best of both the physical science world and the data science world, has really hard problems, works with big companies and has big contracts, and is a team where more than half the people have technical graduate degrees.

This article is spot on. Using AI and ML in an enterprise setting is about changing behavior and helping people to understand why the AI/ML system is making the recommendations (or whatever the output is) that it is. I have seen this in dozens of companies now: even great AI requires cultural change to succeed. And good cultural change can make only mediocre AI models into game changers.