Google AI Experiments – An Online Hands-On Activity

AI (Artificial Intelligence) Experiments with Google are similar to mini games that showcase how AI can easily memorize patterns that it is coded to memorize and recognize these patterns very quickly. This is known as machine learning. For example, you can draw something and have the AI guess it for you (quite quickly) or you can listen to 14000 different sounds…or 14000 different bird sounds!

Our volunteer, Youssef, showcases this with a couple of experiments. He will show you how to access and try them, and even make one yourselves – just watch the short video guide below!

Check out this video which will get you trying multiple AI experiments all through your laptop or mobile phone!

What’s the science behind this?

The main objective of AI is to provide machines with intelligence and the ability to learn, much like humans. How do you get a computer to think like a human? You design the code just like a brain. This is done using artificial neural networks (fancy term for computer brain). Artificial neural networks are split into layers of code that interact with one another, and each layer is split into ‘neurons’ called nodes. After designing the program’s brain, you then need to teach it, which would require a gigantic amount of data. The idea is that if you give the neural network enough data, it’ll be able to do all sorts of things we can like identify faces, handwriting, and all sorts of cool stuff. We only know how to do all that because we’ve been exposed to these things all our lives, and because our neural networks are far more complex than any artificial neural network, at least today. To accelerate the computer’s learning curve, you allow it to process large amounts of data stored in data banks. The content of the data would differ depending on what you want the program to do, for starters, if you want the program to be able identify any cat, you’d have to give it hundreds, thousands, or even millions (the more the merrier) of different cats, from different angles, in different settings. After seeing millions of cats, the program should be able to tell if a picture contains a cat with a high degree of precision (though that would depend on the quality of the neural network).

Where can I do these?

Links for the AI Experiments mentioned in the video can be found below – remember there are a wide array of other experiments to try on these websites, make sure let us know which are your favourites in the comments!

~All credit to Google LLC.