Almost by chance, Mark Kaufmann and his father Steve launched a language-learning software company two decades ago, when they wanted to help an immigrant to Canada who was down on his luck.

Steve, who runs the forestry-software company KP Logix, was learning Cantonese by listening to a Chinese radio station at the time. One day, he heard about a Chinese immigrant—a graduate from a top technical university—who had his life savings stolen from him. “So, [my dad] reached out and said, ‘Hey, I heard about this guy, we have a software company, maybe we could try and … help him get on his feet,’” said Mark.

When the man showed up for work, the Kaufmanns discovered that despite his high TOEFL (English as a foreign-language) test score, his English-speaking skills needed practice. “We thought that was an opportunity,” said Mark. “We had our tech people in the office mock up a prototype … a program that functioned the way my dad had always taught himself languages effectively. That was how it started.”

Two decades later, their West Vancouver-based company called LingQ (pronounced “link”) runs a mobile and desktop platform that teaches its users over 40 languages. In December, it went through another key iteration, with the launch of LingQ version 5.0 that uses artificial intelligence and machine learning to sustain learner interest and convert online media into teaching material.

Mark said a challenge for LingQ is finding ways to keep users motivated. For example, every January, the company has a glut of signups from people keen to get started on New Year’s resolutions, but lose interest shortly after. “Like gym memberships; people have started out with the best of intentions, but not everybody stays with it,” said Mark. 

LingQ lets users import text, audio and YouTube videos into their platform. The software can tell between words that are new, known or tricky for the individual. The system uses that data to analyze imported content and suggest whether it’s right for the learner.

“We can say, ‘Hey, this content has 50 per cent new words. … That’s probably too hard for you.’ If it’s 10 or 15 per cent, that’s probably your sweet spot. Two per cent is probably too low; it’s not efficient for you,” said Mark. “So, in addition to providing statistics for motivational purposes, it allows us to analyze … and optimally recommend content for users at the right level.”

Another feature of version 5.0 is the ability to match audio to transcripts with timestamps. Previously, the manual process was fine if there were only 20 lessons, for example, said Mark. But he wanted to give users access to more content on the internet while maintaining the ability to slice audio into sentence-sized bites.

“Ten years ago, people were asking us for this and we’re like, ‘Well, we can’t possibly do that, we don’t have the people to manually do that.’ But now it’s automatic,” he said.

Mark is already looking into the potential of AI chatbots like ChatGPT for a future version of LingQ to allow learners to practice conversations without needing to schedule a lesson with an instructor. “You can practice and the responses are realistic enough for a language learner. I think there’s a ton of opportunity there,” he said.

Steve’s philosophy on language learning was that the existing classroom approach isn’t as effective as learning from content used by native speakers, said Mark. Now 77, Steve speaks 20 languages to varying degrees of proficiency. He learned French, German, Spanish and Italian while studying and hitchhiking in Europe; and Mandarin and Japanese while working at the Canadian embassies in Hong Kong and Japan.

Mark was a professional hockey player in Japan, Switzerland, Italy, the U.S. and Canada. His sports career included his alma mater Yale University, the Portland Pirates and Team Canada. He laughed at the idea that a hitchhiker and a hockey player started an edtech company.

LingQ has grown from a father-son venture to 23 employees today. “That I’d be here 20 years later, to be where we are—I don’t know what we thought when we started. Maybe we didn’t think that much; we just kind of did it,” said Mark.

The company has so far been bootstrapped with funding from the Kaufmann family’s other forestry-related ventures, KP Wood and KP Logix. “We are fortunate enough to be able to continue to invest,” said Mark.

During the lockdown, people rushed to online learning and activity on LingQ spiked. “People were looking for things to do,” said Mark. “You can only play video games and watch Netflix for so long. You want to try and find something that’s going to give you a leg up and exercise your brain.” But activity has returned to normal since restrictions lifted.

Edtech investment rose, too. Total capital invested in the sector rose from US$18.73 billion in 2019 to US$62.57 billion in 2021, but fell to US$26.99 billion in 2022 as overall tech fundraising slowed, according to PitchBook data.

While the Kaufmanns are already implementing AI in LingQ to recognize language-learning patterns, some argue there’s a need to look at investor hype around AI in education with more skepticism. 

“I’m very cautious,” said Beth Porter, co-founder and CEO of the AI-powered executive-education company Esme Learning. She was previously a vice-president of edX, an online open course provider created by MIT and Harvard. “There were lots of people who’ve made lots of promises about robot tutors,” she said. “We take a much more nuanced view of the world.”

“There’s going to be a need for more proof and less hyperbole around how AI is used in education,” she said.

Despite seasonal and pandemic-induced fluctuations, Mark remains optimistic for the future of edtech. “New stuff comes out all the time and we’ve had to evolve over the years,” he said. “There’s potential in … AI opportunities allowing us to offer more and more content.”