Maths listens and watches

Russell Ivanovic is the chief product officer at Pocket Casts – a premium podcast app that boasts unique audio-enhancement tools thanks to complex algorithms. The app scans audio files for gaps and trims them, then speeds up super-slow voices without you noticing.

“When you speed up someone’s voice, it’ll increase in pitch so they’ll sound like a chipmunk. By implementing a time/pitch algorithm we can speed up someone’s voice, without altering the pitch of the actual voice. Good luck figuring out that one without maths,” says Russell.

Russell’s excited by the prospects of what machine learning could do in the future. Take the recent developments in image recognition: “If I show you a picture, you can immediately tell me if it contains a hotdog or not. A machine 20 years ago couldn’t,” says Russell. Now, artificial intelligence (AI) can trawl through hundreds of thousands of hotdog pictures and use that data to identify a hotdog with pretty good accuracy.

“It may seem simple, but even the simplest software can involve thousands of branching conditions and state machines. A good knowledge of math is essential to reason about these, implement and fix them,” he says.

Maths entertains us

It’s easy to forget how big the internet is. Take YouTube for example; there are over 1.3 billion users on YouTube, uploading 300 hours of video content every minute. Sorting through truckloads of spam and irrelevant content to serve up videos you’ll love requires a pretty sophisticated algorithm.

Video quality and recommendations are calculated in one fell swoop; YouTube’s algorithm takes a channel’s subscriber count, video views and minutes watched to quantify the quality of that content. Then, YouTube will either recommend that channel’s A+ video content to users, or bury their clickbait deep within the pages of a search. The algorithm is working. Reportedly, 70% of YouTube’s viewing time comes from recommended viewing!

Maths plays matchmaker

How can maths predict perfect pairings of people in love, friendship or business like some apps claim to do?

In the 1960s, two American economists pondered a riddle in match-making: with a data set of X number of women and X number of men who have each ranked their potential partners in order of preference, could you pair them successfully so that no couple preferred another partner? The economists, David Gale and Lloyd Shapley, formed an algorithm that made the pairings possible. It’s called the Gale-Shapley algorithm, and it’s got plenty of surprising applications beyond dating apps.

Forty years after the algorithm’s inception, in 2004, economist Alvin Roth decided to test it on U.S. patients in need of kidney transplants against possible donors. In 2003, merely 19 life-saving transplants had taken place. In 2012, Roth and Shapley received the Nobel Memorial Prize for Economic Sciences for matching more than 2000 successful kidney transplants using the matchmaking maths.