How machine learning works, as our CTO explained it to kindergarteners

Machine learning has become a buzzword in tech, so this post explains how it works the way our CTO Albert Padin did. Albert, one of our founders and a Google Developer Expert (GDE) for Machine Learning, was once asked to talk about machine learning to a group of kindergarten students in Singapore, and this is how he explained it.

You know what a lion looks like.

You also know what an elephant looks like.

You know this because you've seen examples of a lion and of an elephant, in real life at the zoo, in movies or in photos on the internet. You recognize an animal once you have been shown what it looks like.

A machine learning model learns the same way. When you give a machine examples (data), it learns over time.

If you show a machine lots of pictures of lions and tell it these are lions, it will then be able to tell what a lion is. If you show it pictures of elephants and tell it these are elephants, it can then identify elephants on its own.

Gaming works the same way. A gamer who has died many times on the same boss battle, against the same repeating attack patterns, will eventually know how to counter and defeat the boss.

Humans need to sleep, though, and we forget things sometimes. Computers have much more memory (which can always be upgraded), they don't forget and they don't need to sleep. In 5 minutes, computers can process data that humans take 20 years to gather.

That is the advantage machine learning has over our own mental processing. A machine learning model processes data as you feed it in, and then it can start identifying and processing similar data on its own right away.

This is why companies around the world are racing to use machine learning to improve business performance.

Now you're ready to explain to your friends, and to preschoolers, how machine learning works. robot wink

Read the story of how machine learning saved us 4 months of work to classify 200,000+ images