Algorithms play role in so much of what we see online and in our day-to-day lives, helping out with everything from setting bail to finding recipes. But while the algorithms of the past were painstakingly coded by humans, the algorithms of the future will be built by robots. They'll be better, more efficient, but also nearly impossible for humans to understand.

Machine learning algorithms don't know anything when they're first created. They only learn how to do anything by stumbling around in the dark, poking at data human researchers give them and doing thousands and thousands of tests to see if they've happened to come up with a valid way to do their simple jobs, something like correctly identifying a human smile or telling the difference between a dog over a cat.

YouTube channel CGP Press takes a look at the basic principle behind how these robot brains learn, and how they learn to learn in the first place.

Machine learning is all about giant datasets and repetition, which why is researchers are so enthusiastic about collecting as much information as humanly possible so they can use it to train their new algorithm-writing bots to write better code than ever.

But hopefully, at the end of the endless testing, the new robot brain has figured out complex questions at speeds better than a human could manage. That's both exciting and somewhat terrifying.