Baidu’s ‘Medical Robot’: Chinese Search Engine Reveals Its AI for Health

On a recent trip to Beijing, Wei Fan, a researcher for Baidu in Sunnyvale, Calif., had to deal with his mother’s unexpected ailment. Her knee ached. The wait to see doctors in the city is often arduous, so Wei called up an old friend who volunteered to stand in line for his mother. The friend waited for more than two hours.

Even after leaving the waiting room, Wei said, patients like his mother may wait longer to see the right physician.

His frustration with China’s overburdened health care system informed Baidu’s latest product: A voice translation app akin to WebMD. Users rattle off a list of symptoms, such as achy joints, red eyes and a cough, and the Chinese search giant sends an immediate diagnostic suggestion (flu, 75 percent odds). Then it links users to a nearby medical specialist.

A majority of Chinese online turn to the Web first for health information, and voice search is far less cumbersome than text, Wei said.

“From a patient’s point of view, you’d rather have something like natural language — something you can talk to, [so] you can describe multiple symptoms at the same time,” he told Re/code. “Our long-term goal is to build a medical robot.”

Wei’s project, called AskADoctor in English, is one of the earliest to emerge from Baidu’s deep-learning division since it hired Andrew Ng, a renowned data scientist and former marquee researcher at Google. And it’s an example of the unique tech interface the company can produce given its privileged access to the world’s biggest nation, which has kept Silicon Valley giants at arm’s length.

The initiative is also another sign of the broader industry trend of tech firms storming into medical sciences with their artificial intelligence guns drawn. Earlier this week, IBM announced plans to acquire medical imaging company Merge Health, turning its data over to IBM’s supercomputers. Google, while not fully public about its medical programs, has similar ambitions. Apple has its wearable health strategy.

Baidu’s advantage comes with scale. Ng’s team talks of delivering research that has a direct impact on the company’s bottom line, reaching “hundreds of millions” of users. China’s tremendous Internet population makes the latter goal easier.

The team is also betting big on voice, a field where it may advance more in China than other rivals. Since February, Baidu’s deep-learning stateside team, around 40 researchers, has worked on building artificial neural networks to process Mandarin. The technique allows machines to render the language — a complicated one, as it’s tonal and character-rich — with far more computing power. (See here for an explanation of neural nets and how tech giants are deploying them.)

“Just as the invention of the touchscreen transformed how we all interact with technology,” Ng said, “I think there’s a potential of speech to make a huge transformation.”

With AskADoctor, the computer voice translation couples with another deep-learning model that ropes in the health data owned and scraped by Baidu across the Chinese Web. Wei said the product can assess 520 different diseases, representing upward of 90 percent of the most common medical problems nationwide. A desktop version is now available, and Baidu plans to release the mobile app soon. Over time, Wei added, Baidu hopes to tie the product in with medical records in China, which are currently in the early stages of going digital.

The product fits with the company’s new focus on connecting online users to offline services — eventually, it will take a cut when it connects users to local doctors. It’s a necessary pivot, an attempt to reinsert the search engine’s relevancy as app usage outpaces the mobile Web. But it’s a costly one: Last quarter, Baidu reported revenue of $2.7 billion, below expectations, and said it plans to invest $3.2 billion in online-to-offline services.

Ng’s AI has helped counteract those costs, according to Baidu. A computer vision-driven improvement to an image product for advertisers improved click-through and paid-click rates, the company said on its earnings call.

The AI team has also brought a headache. In June, Baidu was barred from an international AI competition, in which companies like Google, Facebook and Microsoft compete, for breaking the rules with its image-recognition tech. Ng led the prompt move to fire Ren Wu, the researcher Baidu faulted for the breach, but the incident has damaged the company’s standing in the insular research world.

Baidu did not comment much on the episode, beyond that it had let go of the staff responsible.

Asked what sets Baidu’s AI division apart, Ng returned to size, and not just China’s. Baidu is investing heavily in AI hardware — it clusters large numbers of graphics processing units trained on speech models — something Ng may not have had at Google, which tends to favor a more dispensable approach to hardware.

“I’m pretty confident we’re building supercomputers that let us scale with these deep-learning algorithms bigger, faster than anyone else,” he said.

This article originally appeared on Recode.net.