I once considered teaching myself to make neural networks with Python. I wanted to gather as much data as possible on NBA statistics so that I could win Warren

I once considered teaching myself to make neural networks with Python. I wanted to gather as much data as possible on NBA statistics so that I could win Warren Buffett's $1 Billion dollar challenge to fill out a perfect March Madness bracket. I then did some research, and found that other people have done that before, and that no matter how many perceptrons you have, and no matter how many layers you create, and no matter how much NBA data you download, you can never have more than 70% accuracy in predicting the outcome of NBA games. This stands in stark contrast with the likes of Eliezer Yudkowsky and Ray Kurzweil who claim that enough perceptrons inside an AI will cause that AI to "escape the box" and turn all of humanity into paperclips.