Taking the drudgery out of app testing with Apptest.ai’s Jaejun Hwang
Testing can be the grunt work of app development. Inevitably, quality assurance testers — who are often app developers or even company leaders in early-stage startups — miss bugs. Issues remain for users to uncover, which can potentially drive down app-store ratings.
Jaejun Hwang says he has developed a solution in the form of artificial intelligence (AI)-enabled testing for mobile apps. The CEO and founder of Apptest.ai explains how it works in this week’s edition of the What’s NEXT podcast.
“Not many people are testing,” Jaejun says. “They don’t have enough money, enough time. But AI could change that, thanks to testing that can be done in half an hour compared to many hours by hand. “We found new tricks so that robotic tests become better,” Jaejun says.
The testing challenge
Done properly, testing could take as much time as developing an application. That means, he said, “If you conquer testing, you’re conquering 50 percent of software development,” Jaejun explains.
Yet aside from the time required, the tedium of testing is reason enough for developers to automate the process as much as possible. The need to test on multiple types of devices further adds to the testing challenge, particularly for smaller companies.
In that environment, an AI-driven testing process can help. Otherwise, Jaejun says, “You kind of have to pray that your app’s not going to break down.”
The AI solution
To use Apptest.ai, developers upload their code to the cloud-based service. From there, it gets installed on physical mobile devices for testing by the AI for however long a developer specifies. Apptest.ai then generates a bug report.
The company decided it was best to install code on actual devices rather than software that simulates those devices. “We were thinking about using emulators,” Jaejun says, “but you know, nobody buys that. They want it resolved within real devices.”
The Apptest.ai solution puts an AI program to work on an app to be tested, clicking buttons to make sure they all work, in an approach known as explorer testing. “Maybe he’s not that smart,” Jaejun says. “Maybe it doesn’t even know what to test at all. But if I can make a test bot explore the buttons or input fields as much as possible so that he will find new screens and cover as many pages as possible, that’s a good start.”
The test bot doesn’t have to cover every possible combination of button-presses to catch bugs—nor could it. Instead, the AI program specializes in ferreting out bugs that human testers tend to miss.”If you multiply all of them out, then people say that that’s more than the number of dust grains in the universe,” Jaejun says. “So there’s no way you can test them all.”
To prove his point about buggy apps, Jaejun and his team downloaded the top 100 most popular apps from the Google Play Store and ran them through Apptest.ai. They found that 15 percent of them had at least one bug. “That’s really remarkable when you think about it because these popular apps are usually making a lot of revenue,” he says. “They probably went through a lot of tests. They didn’t pick it up. The AI test bot picked it up.”
Those post-production errors cost more to fix than bugs caught during the development process, Jaejun said, so the AI can help save money on the front end. It’s the kind of problem he’s been working to solve over his 18-year career in computer science.
Disrupting rather than being disrupted
After earning an undergraduate degree in computer engineering from the University of Michigan, Jaejun moved to New York City. There he earned two back-to-back master’s degrees, one in computer science, the other in medical informatics, and started on a Ph.D.
“Then I got to see machine learning and deep learning, and I freaked out,” Jaejun says. What had him worried was the idea that the profession he was training for — computer science — would be rendered obsolete by AI. So he decided to embrace the disruption that threatened to overtake his training.
He settled on an area that seemed to have the widest possible utility to every industry that uses software, which is app testing. He launched Apptest.ai in 2016, got seed money the following year, and by 2019 had attracted investment by Samsung NEXT. Now developers all over the world depend on Apptest.ai to run automated tests on their mobile apps.
More to learn
The AI behind Apptest.ai is just getting started. “Our AI test bot, when compared to the human, he’s like a five-year-old. It needs to go to elementary school,” Jaejun says.
Asked what he’s been doing when he isn’t immersed in AI algorithms, Jaejun says he likes digging in the dirt. “I’ve been a weekend farmer for more than 10 years,” he says. “It’s very relaxing. So if my company fails to become a unicorn, I would start doing farming.”
But so far the financial harvest from Apptest.ai’s success in the marketplace suggests that Jaejun will be dividing his time between farming and coding.