ARTICULAR Public Engagement

Beat the Machine at Its Own Game

Medicines have to be made exactly right before people can take them. To help us, we use computers to monitor all the particles that we make and measure their sizes and shapes. This sounds great, but how does a computer know what a particle looks like? In fact, how do computers know what anything – people, faces, cars, traffic lights – look like?

It turns out that computers can be “trained” to do all sorts of tasks using a process called machine learning. This sounds complicated, but it’s actually pretty much how people learn: just keep trying it until you get it right!

Here, we have trained a computer to detect images of some of our particles. However, we also have some convincing fake images! The computer is pretty good at recognising the difference, but not perfect. We know you have never seen these images before, but can you beat the machine at detecting the fakes?

Play the game at: https://explorathon.shinyapps.io/explorathon2021_btmgame/

(Please note: the game operates best on a Windows PC or laptop. Playing on other platforms may reduce functionality.)

Hopefully, you have now tried the game and managed to beat the machine! Isn’t it interesting that you did not need to know anything about the particle images? You didn’t need to. Simply telling you the answer each time you guessed was enough to help you learn for the next time. This is exactly how computers can be made to do amazing things like:

Facial recognition – phone cameras and photo filters
Driverless cars – detect road signs and other cars on the road
Medical testing – spot anomalies in radiology images
Natural language processing – speech recognition (e.g. Alexa, Siri, Google Assistant)

… even though they do not know anything about the world around them, and don’t even have a brain!

 

 

Beat the Machine at Its Own Game activity was presented at Explorathon ’21 online event:

https://www.explorathon.co.uk/activities/beat-the-machine-at-its-own-game/