[Herald Interview] ‘Korea trails far behind in artificial intelligence’

By Shin Ji-hye

Hanyang University develops artificial intelligence speaker with embedded server and voice recognition only for registered users

  • Published : Dec 6, 2017 - 17:50
  • Updated : Dec 6, 2017 - 17:50
Artificial intelligence may become a threat in the future, but it is not something Korea should be wary of at least for now as the hardware powerhouse belatedly moving into the AI industry trails far behind its rivals, according to a local professor specializing in AI-powered voice recognition technologies.

“AI can sure be a threat to humans in the future by, for instance, taking their jobs. But Korea is now in no position to be concerned by such an issue,” said Chang Joon-hyuk, a professor at Hanyang University’s engineering college, in an interview with The Korea Herald.

“What we have to do now is to give the industry a jump-start by advancing research, development and commercialization to catch up with rival nations,” he added.

Chang is one of a handful of professors specializing in AI voice recognition technologies in Korea. He is the first local professor who developed AI speaker. Last year, he developed voice communication technologies using deep learning for the first time in the world and his research was published in IEEE Transactions on Audio Speech and Language Processing.

The first local AI speaker for South Korea, Nugu, was developed by SK Telecom in September last year, to be followed by KT, Naver and Kakao. Samsung Electronics’ voice-recognition Bixby came out in the market in April this year. 

Professor Chang Joon-hyuk (Park Hyun-koo/The Korea Herald)

They are years behind the global players, Amazon, Google and Apple. The Amazon Echo is now sweeping the artificial intelligence speaker market, with its first launch in 2015. Apple’s HomePod, unveiled in June of this year, is powered by Siri, which was first integrated into the iPhone 4S in 2011.

“It still remains to be seen who will take the leadership but it depends on who can better satisfy customers with voice recognition capabilities and service quality,” he said.

Despite being late, Korea’s three tech players -- hardware maker, mobile carriers and search engine -- will inevitably continue to advance their AI speakers by maximizing their capabilities, Chang said.

“Because voice is the most natural interface in controlling devices, AI speakers will become a hub in the future smart home and take the initiative in the market by collecting all the data and information generated within home,” the professor said. 

Professor Chang Joon-hyuk (Park Hyun-koo/The Korea Herald)

He predicted the usage of AI-powered voice recognition technologies will be expanded into other areas, including robots and automobiles in the future.

Chang’s team at Hanyang University also developed its own AI speaker called Pluto late last month with features of embedded server and voice recognition capability for different users. It took more than four years to develop the device with an investment of around 1 billion won ($910,000) from the government and tech firms.

“Pluto was developed with the aim of sharing its original technologies with local tech firms and possibly commercializing it with additional features such as health care services,” the AI expert said.

The key differentiation of Pluto is “its embedded server that makes user voice data safer and how the device can respond to users faster without concern of network delay. It can also work without internet connection.”

Most AI speakers on the other hand have servers in the cloud environment so all of the users’ voice is delivered first to the firms’ servers before returning to the device to respond to users. This may fuel concerns on the safety of voice data and slow response due to network delay.

The embedded server can be possible because Pluto can understand only 200,000 words compared to other AI speakers that can understand almost unlimited words on a cloud environment. But, Chang said 200,000 words are enough for users to communicate at home.

Pluto’s standout element is its ability to recognize differences in human voices so that the device can work for only registered users. Recognizing a user’s voice is becoming more important as the increasingly smarter AI speaker can now order goods by voice command.

Early this year, when a TV anchor in the US delivered a story of “ordering a dollhouse,” a number of Amazon Echo units in US homes tuned in to the broadcast simultaneously ordered the item. Pluto’s user voice recognition capabilities can prevent such happenings, the professor explained.

Pluto will also be embedded in automobiles in the near future that are becoming smarter.

“We are currently working with a large Korean automaker to embed the Pluto technology in its infotainment system. The embedded AI speaker in a car will possibly be unveiled late next year,” he said, without disclosing the name of the automaker due to the sensitivity of the issue.

Chang has published around 90 papers listed in Science Citation Index to date, and won a Young Professional Award from the Institute of Electrical and Electronics Engineers.

By Shin Ji-hye (