>>717775887
Research of complex systems, since this is where it started (so protein folding)
Due to the the law of large numbers, when you have millions of pieces of data, any form of something being out of the norm starts disappearing.
So you use i.e. AI to simulate a million agents in a traffic simulation since you now don't just simulate the intersection you want to change but an entire city like in Cities Skylines, but you actually DO simulate every single car on there and it has an actual AI with needs, goals, etc. Is every individual agent perfect? No, individually they are all rather flawed. But combine them all and you get a behavior that near perfectly simulates the large interactive systems of an entire traffic network better than ever before.
It's a big reason the push against car dependency became so hot right now. We always knew car dependency was shit, but now we have the simulations to prove that more roads and more lanes can actually DECREASE traffic flow and having more options for cyclists, public transport, etc. can increase traffic flow for the remaining car drivers, even if their roads are now much smaller with fewer lanes.
That's basically what complex systems research is. You have a lot of individual elements that are individually not very complex. But all of these systems combined THAT becomes complex and then it also interacts with a million or more different instances which also changes the outcome. That is where AI shines because that is what machine learning was actually invented for. Every other use case so far is a happy accident in comparison.
