There is a growing body of research surrounding deep reinforcement learning that would reject the idea that

"[ neural networks ] are not a good fit for most real-world problems, where you have to remember something that happened in the past."