Anyone who's come here from the technical side of the internet likely knows of me via my pyaixi project.

This is an implementation of a machine learning algorithm called AIXI.

And while I give a very thorough and technical description of the background and motivation in the README there (as well as kudos to the clever minds who came up with the algorithm), I want to try to convey to a non-technical audience what this is about.

AIXI is an attempt to build the kind of general artificial intelligence (or AI for short) that science fiction is filled with.

It's not the most powerful approach to this - in fact, you can almost always find an algorithm that outperforms it for a certain type of problem or environment.

But where it shines is that it's a universal artificial intelligence algorithm - in that, it can always learn a little, always optimise itself a little, always do a little better. In any environment.

So it's theoretically capable of learning anything. Given enough time, and enough resources.

Unsupervised, so long as the environment gives it some clues on what's good and what's not.

As such, it's a fascinating tool in a broader machine learning toolkit.