Skip to Content

How AI Is Feeding China’s Internet Dragon

China’s biggest Internet company, Baidu, is pushing an ambitious effort to add artificial intelligence to its products.
March 28, 2016

Shortly after walking through the front doors of Baidu in Beijing last November, I was surprised to notice that my face had transformed into that of a cheerful-­looking little dog. As I chatted with one of Baidu’s AI researchers, the version of me shown on his smartphone had sprouted a very realistic-looking wet snout, fluffy ears, and a big pink tongue.

The trick was performed on an app called Face You, released by Baidu last Halloween, which lets you add all sorts of spooky effects or animal characteristics to a digital image of your face. Face You makes use of an AI technique called deep learning to automatically identify key points on a person’s face, so that software can then position and stretch a virtual mask with amazing accuracy.

Deep learning is driving a lot more than just goofy apps at Baidu, though. It is making existing products smarter and helping the company’s engineers dream up many entirely new ideas.

Baidu is one of several companies—Facebook, Google, Microsoft, and IBM among them—that are using deep learning for new applications, such as training computers to hold a conversation, while “migrating older applications from ‘conventional’ machine learning to deep learning,” says Yann LeCun, a professor at New York University, one of the key figures in the history of deep learning, and the director of AI research at Facebook. This is occurring at an impressive pace because deep learning has proved so effective at recognizing patterns and making predictions from data.

Baidu headquarters, Beijing.

Deep learning is essentially an especially effective type of machine learning, a way of having computers program themselves after gobbling up vast quantities of data. It involves feeding data to a large network of simulated neurons, which then gradually learn to recognize abstract patterns in that input. A trained network can then, for instance, spot objects in an image, or determine whether a new e-mail message is legitimate or spam.

The technique is certainly helping Baidu maintain its reputation as one of China’s most innovative home-grown enterprises. By just about any measure, Baidu is China’s most successful Internet business: over 92 percent of the country’s more than 536 million Internet search users employ its portal services and mobile apps. And it continues to grow. In the past year it has moved into new areas, including music streaming, insurance, and banking.

Baidu is using AI to move quickly in a dynamic and competitive tech landscape, says Andrew Ng, the company’s chief scientist and a prominent machine-learning expert. Two years ago the company created an internal group called the Institute of Deep Learning to explore ways to apply the technology across the company. Since then, Ng says, deep learning has helped transform its ad system, significantly improving revenue, and powered all-new efforts like the autonomous driving system Baidu demonstrated in December. “The number of deep-learning applications grows each day,” he says. 

Baidu’s Deep Learning Institute often collaborates on AI projects with a Baidu group located in the United States, called the Silicon Valley AI Lab. The U.S. lab was opened in 2013 to attract AI talent already working or studying in the area.

One of the first things Baidu did after setting up the institute, Ng says, was create a deep-learning platform called Paddle that engineers in other departments could use. And researchers from the institute are often embedded within other departments. As a result, deep learning has been used to improve Baidu’s antivirus filters and to predict when a hard drive in one of the company’s giant server farms will fail, among other things.

As I walked through the institute’s offices at Baidu HQ, evidence of frantic activity and experimentation were all around. The lab recently moved to another building on Baidu’s campus to accommodate a larger team, and dozens of hastily set-up desks were cluttered with boxes yet to be unpacked. On one bookshelf sat a drone, which researchers were using to build 3-D models of street views.

Jiawei Gu, a youthful researcher, demonstrated the face-morphing app, and he said the techniques used to build it might enable the company to branch out into virtual reality. For instance, they could provide a way to bring real-world objects into virtual environments.

Deep learning has been used to improve Baidu’s antivirus filters and to predict when a hard drive in one of its giant server farms will fail, among other things.

Jiawei then showed me a wearable device Baidu has created to help blind or partially sighted people navigate the world. Called DuLight, the device hooks onto your ear like a Bluetooth headset, captures whatever’s directly in front of you, and taps into a deep-learning-based image recognition system to identify it.

Jiawei pointed DuLight at a chair and potted plant, and it said: “Recognizing … light plastic chair … recognizing … lush green potted plant.” When he pointed the device at me, it said: “A man smiles, about 37 years old” (a close guess). He said it could be configured to remember someone’s name and identify that person later.

AI, in the form of deep learning, has already helped improve key Baidu products, including the core search algorithm, by making image search far more accurate. And it has increased the accuracy of the company’s voice recognition engine, which enables voice search as well as a relatively new voice-controlled personal assistant called DuEr. Speech technology could be especially important to Baidu’s future in China, as it offers a more elegant way of using a mobile device than entering Chinese characters on a tiny screen.

“Any company with a lot of data should seriously consider deep learning,” Ng says. “It is a superpower that turns huge amounts of data into huge amounts of value.”

Keep Reading

Most Popular

How a simple circuit could offer an alternative to energy-intensive GPUs

The creative new approach could lead to more energy-efficient machine-learning hardware.

This classic game is taking on climate change

What the New Energies edition of Catan says about climate technology today.

How to opt out of Meta’s AI training

Your posts are a gold mine, especially as companies start to run out of AI training data.

How battery-swap networks are preventing emergency blackouts

When an earthquake rocked Taiwan, hundreds of Gogoro’s battery-swap stations automatically stopped drawing electricity to stabilize the grid.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at with a list of newsletters you’d like to receive.