I Know Kung Fu

and other game-breaking applications of machine learning

Here’s a question for you: Was Neo a highly advanced reinforcement algorithm?

Do you think Neo was hosted on AWS or GCP?

I think that is how he was able to learn Kung Fu by simply observing other people doing it.

Its interesting to re-watch films like this with a modern lens, knowing what I know now about machine learning. You could infer that Neo was the one because he was some game-changing reinforcement learning algorithm. That’s why he was matrix-breaking. This scene looks a lot like the training step of a model.

This realization for me was prompted by a discussion of this paper on the TWIML&AI podcast: Playing hard exploration games by watching YouTube. TLDR; the model learns about the different features in games from watching videos in it’s own time, then is trained to play them by defining reward functions.

This to me sounds a lot like the training we see in The Matrix.

It’s going to be exciting to see what 2019 brings us and if AI will takeover and use humans as an energy source. I think we’re still a few years off.

If you’re interested in the discussion of this paper along with others, check out TWIML talk 217.