Amid Global Race for A.I. Talent, China’s Tencent Sets Up Seattle Lab

They’re still putting the finishing touches on the sixth-floor offices in Bellevue, WA, where Chinese Internet giant Tencent hopes to employ as many as 20 artificial intelligence researchers by next year.

Like technology companies the world over, Tencent is scrambling to recruit as many of the computer scientists and engineers with skills necessary to advance research and build applications with increasingly human-like capabilities, such as recognizing speech and objects in the real world.

Its new office here in the Seattle area, a global center of computer speech research and commercialization, is Tencent’s bid to capture more of the talent attracted by industry leaders including Amazon and Microsoft.

By Tencent’s own estimate, there may be only 300,000 A.I. experts in the world today—up to a third of whom are still in training—while demand for these individuals is in the millions. The company’s Tencent Research Institute released a report earlier this month quantifying the shortfall, as summarized by The Verge.

Meanwhile, the Chinese government has made A.I. technology leadership a national priority. In November, the Chinese government tapped Tencent, along with two of China’s other Internet titans, Baidu and Alibaba Group, for a “national team” to build “open innovation platforms” in various A.I. fields, according to a report in the South China Morning Post, citing a Chinese Ministry of Science and Technology posting.

In Bellevue, a tech hub just east of Seattle, Baidu opened an office in October after acquiring startup company KITT.AI, a spinout from the Allen Institute for Artificial Intelligence. Alibaba, which established a Seattle outpost in 2014, moved to Bellevue last year.

Tencent’s A.I. lab in Bellevue will focus on speech processing, synthesis, speaker identification, and dialogue systems, as well as natural language understanding and generation, says Dong Yu, pictured above, an accomplished computer speech researcher who heads the lab.

Tencent hired Yu away from Microsoft Research, where he was part of a team that built a system that has equaled the human error rate on a benchmark speech transcription task.

Yu says he was attracted by Tencent’s enormous scale and the opportunity to build A.I. systems that could be used by applications such as WeChat, the company’s messaging service, which has some 980 million monthly active users.

Despite the differences between the Chinese and English languages, the underlying technologies used for speech recognition “are almost the same across different languages and it’s even possible that you can use one system to recognize speech spoken [in] different languages,” Yu says.

The Tencent lab in Bellevue will focus on both applied and fundamental research in computer speech.

It also recently began to focus on applications in medicine—helping patients schedule visits and capturing conversations between doctor and patient, Yu says. As in the U.S., China has strong protections on patient data and privacy, Yu says, meaning access to data for training A.I. systems for medical applications can be difficult.

“Typically we will work with several hospitals first, and [there is] very strong enforcement on the privacy issues, so typically the system will be trained on site,” Yu says.

Asked if Tencent, as it establishes this new lab, is putting in place any processes or directives for researchers regarding privacy, ethical, and safety issues raised by advancements in A.I., Yu says:

“Because our main focus here in the Seattle area is mainly speech processing and natural language processing, those are probably less dangerous than other A.I. fields so we don’t have a special team working on the ethical part for these two research areas.”

Between here and Tencent’s main hub of A.I. research in Shenzhen, China, the company employs about 70 A.I. researchers and 300 research engineers, Yu says.

In addition to speech and language processing, Tencent’s A.I. focus areas are basic machine learning theory and computer vision.

Trending on Xconomy