How data science helps machines think a bit more like us
As a kid, you might, like me, have watched the Japanese manga series Doraemon, revolving around a robotic cat that travels back in time from the 22nd century to aid a boy named Nobita. I remember wondering if the magical gadgets Doraemon produces could really exist.
I also remember looking at what technology did exist – the cars, televisions and phones surrounding us – and thinking, “Why can’t Thailand create these such things for itself?”
Those two questions inspired me to study telecommunications engineering. I completed an undergraduate degree from King Mongkut’s Institute of Technology Lad Krabang. Then, I jumped into the AI world by pursuing a masters degree at Germany’s Ulm university followed by a doctorate at Caen University, in France, majoring in computer visions and data mining.
My first job was linked to the future of self-driven cars. Working at Stuttgart-based automotive company, Daimler AG, I researched ways for cars to detect pedestrians. To do that, we used what’s called artificial intelligence (AI).
AIs grant a machine the capacity to think a little bit more like a human – the ability to read complex situations, make a decision by itself, and even learn, all of which are critical skills for technological advances like self-driving cars.
AIs are created with the idea that the best way to learn is from experience. Humans remember from past experience the best and worst ways to accomplish a particular task. Every time we make a mistake, our brains add that feedback to the learning process in order to prevent those mistakes from happening again.
This auto-feedback system has long been studied by researchers. And we are now applying it to machine learning. We turn that feedback process into a piece of software, an algorithm, which imitates human brain functions.
One such piece of code is Google’s AlphaGo, an AI that recently beat the world’s best human go player. Conceptually, the programer lets two robots play each other a million times. Every time they failed, they learned from that, until their tactic becomes increasingly refined. Ultimately, they reached a point where they can beat the very best humans.
Data science is simply that same method of machine learning applied to business. DTAC, for example, analyses its customers’ data with hundred of variables that affect the decision factors behind the services they use, such as Internet speed, number of calls, price plan,
etc.
It allows DTAC to offer more relevant packages that match their customers’ usage, as opposed to letting the customer sift through thousands of offers in search of the one that best fits them.
Currently, I’m working on an AI-based chatbot (a computer program designed to simulate human conversation with human users) to relieve the call centre team of routine queries and allow them to focus on more complex cases.
They’re clearly doing a great job since DTAC just won The Consumer Protection Thailand Call Centre Award. As for the bot, it’s currently available for SMS queries and will be launched on Facebook next month.
During my free time, I also created “Botnoi,” an experiment I coded to allow users to talk with and befriend a robot. Botnoi has chalked up more than a million downloads through Line and Facebook. Botnoi also received the Line Bot Award last year, from Line Corporation.
Its impact is beyond what I could have imagined. It once helped a suicidal person by just being a good friend to them. Botnoi is still a far cry from Doraemon, but I am proud of the contribution it makes to society as a chatbot friend for all.
Winn Voravuthikunchai is group data scientist at the Telenor Group.