The Paperclip Maximizer

The idea of the paperclip maximizer is you basically have some goal that you want to train an artificial intelligence system on. And when we look at being able to train an AI on a particular goal like winning chess or winning AlphaGo, we see that if they understand the boundaries of how they could optimize that, they get capable of things that humans couldn't do very quickly. So as it gets smarter and actually as it gets more intelligent, it gets better at getting more intelligent. It does whatever goal you train it on plus it gets better with an exponential curve because as it Gets smarter, it's also smarter at figuring out how to get smarter. They're then all applied to doing whatever