The Future of Machine Learning

In 2012 we wrote papers showing that human genomes, where in a certain university class that had been discovered for L one penalized algorithms. In those papers we predicted that in order to solve height, you would need a few hundred thousand labeled individuals. And indeed as soon as that data set of that size became available in 2017, within a month we had solved it and built the hyperdictor using these methods. So now you have to be a real nerd to care about thisYou have to be applied mathematician or someone who actually likes theoretical machine learning. But I would say we deserve a big like prize in Stockholm because we made a non trivial theoretical prediction.