MIT Technology Review
The Algorithm
Artificial intelligence, demystified
The companies that fuel China’s AI ambitions
01.22.19
Hello Algorithm readers,

China’s AI industry lives and dies by BAT. That stands for Baidu, Alibaba, and Tencent, the three Chinese tech giants that are loosely equivalent to Google, Amazon, and Facebook. These three are not just developing and deploying AI themselves. Their deep pockets have also funded a broad range of AI companies, focused on everything from smart cities to finance to education.

Last week, Chinese media outlet Huxiu.com published a graphic that visualizes the full extent of their involvement across China’s AI industry. (The graphic is quite complicated, so I translated it into three bar charts below. Here is also an English translation of the article it originated from, courtesy of Jeffrey Ding, who writes the ChinAI newsletter.) It revealed that BAT invests in 53% of the nation’s 190 major AI companies. This may not surprise those of you who closely follow China’s AI ecosystem. But it’s quite a different topology for those more familiar with Silicon Valley’s.

BAT invests in more AI companies than other other AI giants

Looked at one way, the landscape shows how intense the competition is between the three. While they each have a main pillar of expertise—Alibaba in e-commerce, Tencent in social networking, and Baidu in search and information indexing—they are also challenging one another head-on across dozens of industries.

BAT has competing investments across dozens of industries

Looked at another way, the scale of BAT’s involvement shows just how integral they are to China’s bid to be a global leader in AI by 2030. Their expertise and funding set the direction and pace of the technology’s development, but their weaknesses also affect the robustness of China’s ambitions.

BAT has promoted a top-heavy, application-focused AI industryAs the graphic highlights, BAT’s investments have promoted a top-heavy AI industry: lots of companies dedicated to AI applications with far fewer dedicated to developing the technologies that underpin it, including the algorithms and advanced silicon chips behind the breakthroughs in machine vision, natural language processing, and other AI capabilities.

Experts have warned about this top-heaviness before. China’s astronomical rise in AI leadership is currently buoyed by its abundance of data and lax views on privacy. In the short term, those conditions make it fertile ground for highly profitable machine-learning applications. But the country still lags behind the US in its efforts to expand existing AI capabilities through fundamental research. In the long term, that could place a ceiling on how much China will continue to benefit from the technology’s revolution.


US veterans are helping DeepMind solve its health data shortage. Nearly four million adult Americans are hospitalized each year for acute kidney injury, a life-threatening complication. The US Department of Veterans Affairs (VA), a federal agency that provides healthcare services to military veterans, thinks AI can lower that number. The idea is if software can predict those most at risk of developing the disease, doctors will be better equipped to prevent it from happening.

In collaboration with DeepMind, the VA offered the company 700,000 medical records from US veterans over a 10-year period to help train its algorithms. The records were encrypted and sanitized and remain under the VA’s control. If the algorithms prove effective in making predictions on historical data, the collaborators will likely test the system on live patients at the VA clinic.

The project is a notable example of how AI could transform healthcare—by using machine-learning to identify the patients who'd benefit most from preventative care. But such efforts are usually hindered by a lack of training data because of the strong privacy protections specific to the industry.

Fortunately for DeepMind, the VA has millions of electronic health records, representing one of the most comprehensive collections in the US. It highlights an interesting path forward for AI researchers working in this arena: to partner directly with organizations willing to proffer up their medical data troves.

TR archives


Will Knight, senior AI editor, on China’s ambitions to create an AI chip industry: “iFlytek’s translator shows off AI capabilities that rival those found anywhere in the world. But it also highlights a big hole in China’s plan, unveiled in 2017, to be the world leader in AI by 2030. The algorithms inside were developed by iFlytek, but the hardware—the microchips that bring those algorithms to life—was designed and made elsewhere.” Read more here.

The votes are in

Thank you to everyone who wrote responses to our last issue on the algorithms used in the American criminal justice system. People fell into two camps: those who think machine-learning algorithms should not be used in this context at all and those who think they could be beneficial if applied cautiously—but not in their current form.

As always, you can send your thoughts and questions on any issue to algorithm@technologyreview.com.

Featured EmTech Digital Speaker: Kent Walker

Every industry leader should be considering how to implement ethical AI. Google's SVP of Global Affairs and Chief Legal Officer will join us on stage at EmTech Digital to discuss both policy and implementation and how to establish principles of responsible AI. Secure your ticket today!


Research

Social animal. Humans are instinctively tribal creatures. When we observe the interactions of people around us, we can intuitively infer who we should get along with and who we shouldn’t. This might sound like a negative instinct, but it is actually what makes teamwork possible. Researchers at MIT believe this skill may be an important prerequisite for more social AI, and are now trying to replicate it in a machine.

This idea isn’t totally new. Game-playing AI agents also require knowledge of the social landscape to know who to cooperate and compete with. But they’re given these relationship structures explicitly within the rules of the game, while humans can quickly pick them up in ambiguous situations.

Inspired by this ability, the researchers developed a new machine-learning algorithm to figure out the relationships among multiple agents through a limited number of observations. They then ran two experiments to test the algorithm’s performance. In the first one, it had to infer the alliances of players in a video game by watching several sequences of gameplay. In the second one, it had to predict the players’ actions in the same video game to see whether it truly understood each of the players motivations. It wasn’t trained for either task.

In both experiments, the algorithm’s inferences and predictions closely corresponded to the judgements of humans, demonstrating its ability to rapidly generalize social structures from very little data.

Bits and bytes

Chinese courts are experimenting with AI
The goal is to standardize judgements for similar cases. (The Diplomat)
+ So are courts in the US—a controversy we discussed in last week’s Algorithm. (TR)

People are falling in love with robots and avatars
“Digisexuals” may be a new emergent sexual identity, say futurists. (NYT)

The life insurance industry is using machine learning to attract millennial customers
Companies are aggregating publicly available data to automate the insurance application process. (MarketWatch, video)

Facebook is funding a new AI ethics institute
It will investigate issues around the safety, fairness, privacy and transparency of the technology. (Bloomberg)

AI is coming to help filmmakers
Researchers are developing software to automatically edit raw footage into rough cuts. (WIRED)

Quotable

Subsequent generations will have never known a distinction between their online and offline lives. They may grow up with sex education chatbots, make love to the universe in their own V.R.-created world, or meet their significant other through a hologram.

Bryony Cole, the founder of the media company Future of Sex, on the inevitable blurring of physical and digital love

Karen Hao
Hello! You made it to the bottom. Now that you're here, fancy sending us some feedback? You can also follow me for more AI content and whimsy at @_KarenHao.
Was this forwarded to you, and you’d like to see more?
New Call-to-action
You received this newsletter because you subscribed with the email address:
edit preferences   |   unsubscribe   |   follow us     
Facebook      Twitter      Instagram
MIT Technology Review
One Main Street
Cambridge, MA 02142
TR