It’s more that engineering is hard. Understanding how fusion works is science. Making fusion generate energy profitably is engineering.
There are lots of potentially better battery chemistries. Solid state bateries, Zinc-air batteries, etc. The problem is not coming up with the materials. The problem is getting them to work reliably under all conditions for a long time.
I saw an item that said one restriction on the number of CyberTrucks being built right now is that Tesla is still trying to perfect volume manufacure of dry batteries, which would weigh less that wet ones and less likely to burn.
I would question whether chemistry really is quite as ‘solved’ as this.
Indeed. This seems such a common notion. Chemistry is waaay more complex than people imagine from a bit of high school science. About the most useful grounding is that all chemistry is quantum electrodynamics. Every electron interacts with every other electron according to QED. At least only outer shell electrons tend to matter. But the complexity goes insane before you can blink.
Chemistry is like being led into a totally dark room filled with Lego, with thick gloves on. You know the basic rules. But there is a big gap between knowing how the units click together and being able to make anything interesting, let alone useful. Even if you have plans, there is no guarantee you can make it. Usually there are no plans.
sounds like a great field for an expert-AI to step into … let’s see what happens there over the next few years with this regards…
interesting times we live in, indeed!
There are AIs being applied right now. But to say “AI” is spreading a very wide net. There are huge databases of interesting things that AIs have been let roam over to try to gain some traction. But as the saying goes, garbage in, garbage out. You need really good quality data, and that is problem.
Lots of important chemistry can be reduced to electron densities (at least to a first approximation, assuming nothing moves) and something AI algorithms can be good at is learning how to fit multidimensional data. So you might hope that appropriately trained AIs can outpace brute force fitting algorithms. What is a lot harder is getting novel chemistry. And synthesis is a whole new ball game. As another tool in the practising chemist’s backback there is a lot to hope for.
So indeed, the dogs are loose, and it is going to be interesting.
Still not sure what AI is doing here or why AI speeds up discovery of it.
Absolutely, but in both cases (batteries and fusion energy), there’s still a LOT of science left to be done before it’s strictly an engineering exercise, like say… making a fission power plant or powering something with present-day batteries.
The number of possible chemical combinations is staggering and very time-consuming to test. I can totally see how AI might be able to find something that has promise for the application at-hand so it can be further explored without wasting time on all the dead-ends.
Don’t count out incentives. From the perspective of say 2000, laptops and cell phones provided endless demand for lithium ion, making research a safe bet whatever the success or timing of electrical cars or utility scale demand. Even today, the utility market was only about $50 billion in 2023, though it’s likely to go up 5 times by the end of the decade.
Around 6 manufacturers were expected to launch sodium-ion battery products last year. Sodium ion has shorter battery life, lower energy density, and greater weight than lithium ion, but who cares? It’s cheap and less prone to igniting than lithium-ion, making it good for utilities (and problematic for electric cars, though China is exploring that possibility).
Cite: McKinsey article, from sometime in 2022:
Can you elaborate on this? This is the part I’m confused and don’t really understand.
Maybe an analogy will help. Think of all the ingredients in a typical pantry and all the ingredients in a typical refrigerator. If I ask you to bake a cake, and you’ve never done it before, you’ll probably have no idea what to do. If you have done it before, you’ll probably grab a handful of staple ingredients (flour, eggs, sugar, butter, baking powder, …), preheat the oven, combine wet ingredients, combine dry ingredients, mix wet into dry (taking care not to overmix), butter and flour a pan, add the batter, and place in the oven, check for doneness after maybe 25 or 30 min, take it out, know enough to let it rest before you cut into it, …
Now I ask you to bake cookies. You’ll take the same ingredients, but in different proportions. You might set the oven to a different temperature, you might combine the ingredients in different steps (creaming the butter and sugar, for instance), you’ll use a cookie sheet instead of a cake pan, you’ll put them in the oven for much less time and you’ll be looking for different characteristics to decide if they’re done.
Now make a loaf of bread. The ingredients are similar, but definitely not the same. The preparation steps are completely different. The baking temperature and time are totally different. The way to check if it’s done is different. The treatment after it comes out of the oven is different.
Now, we get to what you’re asking of developers of battery materials. Go bake something that nobody has ever baked before, but don’t use too much butter, because it’s expensive, and don’t use wheat flour because you’re baking for somebody with Celiac, and don’t use salt because somebody else has high blood pressure, and don’t use sugar because a third person is diabetic. Oh, and make sure it’s delicious. And it needs to be shelf stable for at least a month.