DeepMind's 'StarCraft' Victory Was as Worrying as It Was Impressive

I hadn’t caught an analysis of alphastars performance so this was interesting. The consensus in the HN comments seemed to be that alphastar was overall not very impressive, relied on manipulating lots of ui elements really quickly, but machine learning experts were very impressed that this result was achievable at all with modern ML.

We haven’t even filled out the potential of basic computing. It wasn’t until very recently that lots of people started carrying out mainstream culture over the internet. And there are still tons of things that are yet to transition to being done through computers. The tidal wave of stuff that is going to be radically changed by ML is coming. I think people have this idea that it’s already here and look, things are still pretty normal. We haven’t even seen the results of general computing and networking. We haven’t even begun to see the result of ML. It’s going to be unpleasant.