AI Can Read! Tech Firms Race to Smarten Up Thinking Machines

Seven years ago, a computer beat two human quizmasters on a Jeopardy challenge. Ever since, the tech industry has been training its machines even harder to make them better at amassing knowledge and answering questions.

And it’s worked, at least up to a point. Just don’t expect artificial intelligence to spit out a literary analysis of Leo Tolstoy’s War and Peace any time soon.

Research teams at Microsoft and Chinese tech company Alibaba reached what they described as a milestone earlier this month when their AI systems outperformed the estimated human score on a reading comprehension test. It was the latest demonstration of rapid advances that have improved search engines and voice assistants and that are finding broader applications in health care and other fields.

The answers they got wrong — and the test itself — also highlight the limitations of computer intelligence and the difficulty of comparing it directly to human intelligence.

Error! Error!

“We are still a long way from computers being able to read and comprehend general text in the same way that humans can,” said Kevin Scott, Microsoft’s chief technology officer, in a LinkedIn post that also commended the achievement by the company’s Beijing-based researchers.

The test developed at Stanford University demonstrated that, in at least some circumstances, computers can beat humans at quickly “reading” hundreds of Wikipedia entries and coming up with accurate answers to questions about Genghis Khan’s reign or the Apollo space program.

The computers, however, also made mistakes that many people wouldn’t have.

Microsoft, for instance, fumbled an easy football question about which member of the NFL’s Carolina Panthers got the most interceptions in the 2015 season (the correct answer was Kurt Coleman, not Josh Norman). A person’s careful reading of the Wikipedia passage would have discovered the right answer, but the computer tripped up on the word “most” and didn’t understand that seven is bigger than four.

“You need some very simple reasoning here, but the machine cannot get it,” said Jianfeng Gao, of Microsoft’s AI research division.

Human vs. machine

It’s not uncommon for machine-learning competitions to pit the cognitive abilities of computers against humans. Machines first bested people in an image-recognition competition in 2015 and a speech recognition competition last year, although they’re still easily tricked. Computers have also vanquished humans at chess, Pac-Man and the strategy game Go.

And since IBM’s Jeopardy victory in 2011, the tech industry has shifted its efforts to data-intensive methods that seek to not just find factoids, but better comprehend the meaning of multi-sentence passages.

Like the other tests, the Stanford Question Answering Dataset, nicknamed Squad, attracted a rivalry among research institutions and tech firms — with Google, Facebook, Tencent, Samsung and Salesforce also giving it a try.

“Academics love competitions,” said Pranav Rajpurkar, the Stanford doctoral student who helped develop the test. “All these companies and institutions are trying to establish themselves as the leader in AI.”

Limits of understanding

The tech industry’s collection and digitization of huge troves of data, combined with new sets of algorithms and more powerful computing, has helped inject new energy into a machine-learning field that’s been around for more than half a century. But computers are still “far off” from truly understanding what they’re reading, said Michael Littman, a Brown University computer science professor who has tasked computers to solve crossword puzzles.

Computers are getting better at the statistical intuition that allows them to scan text and find what seems relevant, but they still struggle with the logical reasoning that comes naturally to people. (And they are often hopeless when it comes to deciphering the subtle wink-and-nod trickery of a clever puzzle.) Many of the common ways of measuring artificial intelligence are in some ways teaching to the test, Littman said.

“It strikes me for the kind of problem that they’re solving that it’s not possible to do better than people, because people are defining what’s correct,” Littman said of the Stanford benchmark. “The impressive thing here is they met human performance, not that they’ve exceeded it.”

China Online Quiz Craze Lures Prize Seekers, Tech Giants

It seems like a game everyone wins: Some of China’s biggest tech companies, looking to hook in new consumers, are using cash prizes to draw millions of contenders to mobile-based online quiz shows.

Up to 6 million people at a time log into the free, live games on their smartphones to answer a series of rapid-fire questions in an elimination battle, with those remaining sharing the prize money.

Over the weekend, search engine giant Baidu and video game maker NetEase launched their own online shows, joining news feed platform Toutiao, Alibaba Group Holding Ltd-owned UCWeb and Wang Sicong, the scion of Chinese billionaire Wang Jianlin.

But how they will cash in on the games and stay on the right side of government censors might prove to be a tricky question.

The trivia games have drawn some controversy, heightened by a broader crackdown on online content during the last year under President Xi Jinping, from livestreams and blogs to a campaign against internet addiction.

This month, one quiz show, “Millions Winner,” backed by internet security company Qihoo 360, apologized after it was chastised by a regulator for listing Taiwan and Hong Kong, over which China claims sovereignty, as independent countries in a question.

How firms will monetize the craze is also not yet clear, though some companies, such as online retailer JD.com, have already jumped on the trend, sponsoring shows to help raise their profiles. Many of the games show ads to players during the shows.

“If you ask me why I do this, to be honest, I don’t really know if I can make money,” Zhou Hongyi, chairman of Qihoo, said at an event where he presented a contestant with a 1 million yuan ($156,115.84) prize check two weeks ago. “But from a user’s perspective, I think this is really fun.”

The quiz mania underlines the fierce appetite of China’s consumers for internet entertainment, a trend helping drive billions of dollars of investment into digital news portals, online gaming, internet advertising and television content.

“I heard about this game from a friend who won 1,700 yuan in one day. I immediately decided to join up myself,” Wang Ting, a 26-year-old graduate student in Qingdao, told Reuters. She now spends three hours each day on her phone playing the games.

Uncertain future

Questions, read by a live host, might include: “From which country were pineapples imported to China in the 16th century?” “In which dynasty was the lamb hot pot invented?” or “How many fingers does Mickey Mouse have?”

Contestants get 10 seconds — a time frame designed cut out cheating — to select the correct answer from a choice of three.

Winnings can be up to 3 million yuan per game, but are often split between many winners.

Toutiao parent Bytedance said that “millions of our users” had taken part in its live quiz “Million Dollar Hero” since the show launched at the start of January. It also has a tougher “Hero Game” with harder questions and bigger prizes.

“We’ve been running for just two weeks, so it’s still in the very early stages, but it’s encouraging to see how the game has taken off across the country, and with all age groups,” the company said in a statement to Reuters.

Toutiao, a highly popular news feed app, was valued at around $20 billion in a fundraising last year, sources close to the company told Reuters.

Raymond Wang, managing partner at Beijing law firm Anli Partners, said the shows were a “relatively low-cost way to get to users,” but cautioned there were political and technical risks.

Wang Ran, a prominent investor and head of Beijing-based private investment bank CEC Capital Group, posed a question on his WeChat account about the future of the online quiz show trend.

“A) Growing numbers will jump into the market. B) Someone will win 10 million yuan in one go. C) Authorities will strictly crack down on it. 10 seconds. Go!”

Social Media Has Mixed Effect on Democracies, Says Facebook

Facebook took a hard look in the mirror with a post Monday questioning the impact of social media on democracies worldwide and saying it has a “moral duty” to understand how it is being used.

Over the past 18 months, the company has faced growing criticism for its limited understanding of how misinformation campaigns and governments are using its service to suppress democracy and make people afraid to speak out.

“I wish I could guarantee that the positives are destined to outweigh the negatives, but I can’t,” wrote Samidh Chakrabarti, Facebook’s product manager of civic engagement.

Since the 2016 U.S. presidential election, Facebook has been looking more critically at how it is being used. Some of what it found raises questions about company’s long-standing position that social media is a force for good in people’s lives.

In December, in a post titled “Is Spending Time on Social Media Bad for Us?” the company wrote about its potential negative effects on people.

The self-criticism campaign extended to Facebook CEO Mark Zuckerberg’s personal goals. Each year he publicly resolves to reach one personal goal, which in the past included learning Mandarin, reading more books and running a mile every day.

This year, Zuckerberg said his goal is to fix some of the tough issues facing Facebook, including “defending against interference by nation states.”

Foreign Interference

During the 2016 U.S. election, Russian-based organizations were able to reach 126 million people in the U.S. with 80,000 posts, essentially using social media as “an information weapon,” wrote Chakrabarti. The company made a series of changes to make politics on its site more transparent, he wrote.

False News

Facebook is trying to combat misinformation campaigns by making it easier to report fake news and to provide more context to the news sources people see on Facebook.

“Even with these countermeasures, the battle will never end,” Chakrabarti wrote.

One of the harder problems to tackle, he said, are so-called “filter bubbles,” people only seeing news and opinion pieces from one point of view. Critics say some social media sites show people only stories they are likely to agree with, which polarizes public opinion.

One obvious solution – showing people the opposite point of view – doesn’t necessarily work, he wrote. Seeing contrarian articles makes people dig in even more to their point of view and create more polarizations, according to many social scientists, Chakrabarti said.

A different approach is showing people additional articles related to the one they are reading.

Reaction to Facebook’s introspection was mixed with some praising the company for looking at its blind spots. But not everyone applauded.

“Facebook is seriously asking this question years too late,” tweeted Jillian York, director for international freedom of expression for the Electronic Frontier Foundation.

Amazon Opens Store With No Cashiers, Lines or Registers

No cashiers, no lines, no registers — this is how Amazon sees the future of in-store shopping.

The online retailer opened its Amazon Go concept store to the public Monday, selling milk, potato chips and other items typically found at a convenience shop. Amazon employees have been testing the store, which is at the bottom floor of the company’s Seattle headquarters, for about a year.

The public opening is another sign that Amazon is serious about expanding its physical presence. It has opened more than a dozen bookstores, taken over space in some Kohl’s department stores and bought Whole Foods last year, giving it 470 grocery stores.

But Amazon Go is unlike its other stores. Shoppers enter by scanning the Amazon Go smartphone app at a turnstile. When they pull an item of the shelf, it’s added to their virtual cart. If the item is placed back on the shelf, it is removed from the virtual cart. Shoppers are charged when they leave the store.

The company says it uses computer vision, machine learning algorithms and sensors to figure out what people are grabbing off its store shelves.

Amazon says families can shop together with just one phone scanning everyone in. Anything they grab from the shelf will also be added to the tab of the person who signed them in. But don’t help out strangers: Amazon warns that grabbing an item from the shelf for someone else means you’ll be charged for it.

At about 1,800 square feet, the store will also sell ready-to-eat breakfasts, lunches and dinners. Items from the Whole Foods 365 brand are also stocked, such as cookies, popcorn and dried fruit.

The company had announced the Amazon Go store in December 2016 and said it would open by early 2017, but it delayed the debut while it worked on the technology and company employees tested it out.

Facebook Should Pay ‘Trusted’ News Publishers Carriage Fee: Murdoch

Media mogul Rupert Murdoch on Monday called on Facebook to pay “trusted” news publishers a carriage fee, similar to the model used by cable companies, amid efforts by the social media company to fight misinformation on its platform.

“Facebook and Google have popularized scurrilous news sources through algorithms that are profitable for these platforms but inherently unreliable,” Murdoch, who controls the Wall Street Journal as executive chairman of News Corp., said in a statement.

Facebook Chief Executive Mark Zuckerberg said on Friday his company would fight misinformation and sensationalism on its platform by using member surveys to identify “trustworthy” outlets.

“There has been much discussion about subscription models but I have yet to see a proposal that truly recognizes the investment in and the social value of professional journalism,” Murdoch said.

The quality of news on Facebook has been called into question after alleged Russian operatives and spammers spread false reports on the site, including during the 2016 U.S. election campaign.

Facebook and Google did not immediately respond to requests for comment.

Star Trek Tricorder Moves Closer to Reality

Assessing someone’s medical status was easy on the TV series, Star Trek. Dr. McCoy just waved his tricorder over the patient, and any broken bones, concussions or internal bleeding were instantly revealed.

While in real life, ultrasounds and x-rays help physicians diagnose everything from breast cancer to kidney stones, those scans can not reveal what is inside the masses.  Having that immediate knowledge could help millions of patients avoid unneeded stress and surgery.  

Purdue University Biomedical Engineering professor Ji-Xin Cheng has devoted his life’s work to technology that will be able to provide that internal view. He and his team have developed several medical tools that help diagnose patients using sound and light. “Eventually we want to make a device like the tricorder in Star Trek,” he explains, “so our dream is to make a movie into a real practice.”

Label-free imaging

In conventional medicine, surgeons must either cut out suspect tissue for analysis, or risk exposing already very sick patients to fluorescent dyes and nanoparticles.  These “labels” light up lesions so doctors can study them.

Team member Jesse Vhang explains their technique – called “label-free imaging” – eliminates more invasive or toxic procedures by bouncing light off molecules in the tumor.

“We do not need a label,” he points out. “We can basically look at the vibrations of the molecules and these vibrations can generate signals in our microscope.”

Those vibrations serve as molecular fingerprints, unique to each type of molecule.  The patterns can be mapped to identify such things as lipids, proteins, and nucleic acids.  In essence, the devices give doctors the ability to look at a patient in three instead of two dimensions.

Vhang says, “For example we could use this to image biological samples from patients.  We can see if the patient has cancer, which usually accumulates a lot of lipids.” The label-free imaging devices have shown promise in identifying kidney, liver, and breast cancer.  

MarginPAT

One device, the MarginPAT, funded by the National Institute of Health, will also help breast cancer surgeons remove tumors more efficiently and accurately.  In the United States, about a quarter of all breast cancer patients must undergo a second surgery to remove missed malignant cells. The developers expect MarginPAT will dramatically reduce that number.

Cheng and his partner, Dr. Pu Wang, founded Vibronix to manufacture the device.  Wang says it could revolutionize medicine around the world.

“I think this will be good in mainland China where medical practice is not as good as the tier one hospitals in the big cities or aboard.  They will be able to use the setup to provide the same surgery as the big city doctors.”

If all goes as planned, the MarginPAT will be on the market within three years and several more of Cheng’s label-free imaging devices will not be far behind, making his dream of a real Star Trek tricorder one step closer to reality. 

US Tests Nuclear Power System to Sustain Astronauts on Mars

Initial tests in Nevada on a compact nuclear power system designed to sustain a long-duration NASA human mission on the inhospitable surface of Mars have been successful and a full-power run is scheduled for March, officials said on Thursday.

National Aeronautics and Space Administration and U.S. Department of Energy officials, at a Las Vegas news conference, detailed the development of the nuclear fission system under NASA’s Kilopower project.

Months-long testing began in November at the energy department’s Nevada National Security Site, with an eye toward providing energy for future astronaut and robotic missions in space and on the surface of Mars, the moon or other solar system destinations.

A key hurdle for any long-term colony on the surface of a planet or moon, as opposed to NASA’s six short lunar surface visits from 1969 to 1972, is possessing a power source strong enough to sustain a base but small and light enough to allow for transport through space.

“Mars is a very difficult environment for power systems, with less sunlight than Earth or the moon, very cold nighttime temperatures, very interesting dust storms that can last weeks and months that engulf the entire planet,” said Steve Jurczyk, associate administrator of NASA’s Space Technology Mission Directorate.

“So Kilopower’s compact size and robustness allows us to deliver multiple units on a single lander to the surface that provides tens of kilowatts of power,” Jurczyk added.

Testing on components of the system, dubbed KRUSTY, has been “greatly successful — the models have predicted very well what has happened, and operations have gone smoothly,” said Dave Poston, chief reactor designer at the Los Alamos National Laboratory.

Officials said a full-power test will be conducted near the middle or end of March, a bit later than originally planned.

NASA’s prototype power system uses a uranium-235 reactor core roughly the size of a paper towel roll.

President Donald Trump in December signed a directive intended to pave the way for a return to the moon, with an eye toward an eventual Mars mission.

Lee Mason, NASA’s principal technologist for power and energy storage, said Mars has been the project’s main focus, noting that a human mission likely would require 40 to 50 kilowatts of power.

The technology could power habitats and life-support systems, enable astronauts to mine resources, recharge rovers and run processing equipment to transform resources such as ice on the planet into oxygen, water and fuel. It could also potentially augment electrically powered spacecraft propulsion systems on missions to the outer planets.

Move Over Traditional Billboards. Make Way for 3D Holographic Ads

Move over traditional billboards. Three-dimensional, slightly hypnotic holograms may soon replace two-dimensional signs and ads. Several companies with this technology said 3D holograms will revolutionize the way businesses and brands talk to potential customers.

“It’s already replacing billboards, LED screens, LCD screens, because there hasn’t been any revolution in the display industry for decades,” said Art Stavenka, founder of Kino-mo, a company with offices in London and Belarus. 

The main hardware of the technology is a blade that emits a strip of light creating holograms of images and words. Multiple blades can be synchronized for larger holograms.

“As soon as this piece of hardware spins, you stop seeing hardware and you start seeing (a) hologram, and the piece of hardware spins fast enough so a human eye does not see any rotation, and it sees the amazing holographic image,” said Stavenka.

Another company developing this type of device is Hologruf, with a presence in both the U.S. and China. 

“In the not so distant future on every street corner, there will be these types of ad displays just like in a science fiction movie,” said Hologruf’s Quan Zhou. 

The applications for 3D holographic displays include shopping centers, train stations and restaurants. 

For franchises such as fast food restaurants that want these displays in more than one location, “they have the capability to manage multiple devices around the world from a central location,” said Hologruf’s co-founder, Ted Meng. 

The cost of a blade ranges anywhere from around $1,300 to just over $3,000, depending on the manufacturer. 

The competition has begun for this technology. Kino-mo has customers in 50 countries on almost every continent. It will be releasing an outdoor version sometime in 2018. Hologruf said it already has a product to replace outdoor billboards.

“We can make it to be water proof, wind proof and work under all kinds of extreme environmental conditions,” said Zhou.

So what would Tokyo or Times Square in New York look like in a few years? Stay tuned.

Traditional Billboards Make Way for 3-D Holographic Ads

Those two-dimensional billboards that dot the landscape of many cities around the world may soon be replaced — with 3-D holograms. Companies working on this technology say it will revolutionize the way businesses and brands talk to potential customers. VOA’s Elizabeth Lee got a glimpse of advertising’s future at the recent Consumer Electronics Show in Las Vegas.

Facebook to Prioritize ‘Trustworthy’ News

Social media giant Facebook said Friday that it would begin to prioritize “trustworthy” news outlets on its site in order to counteract “misinformation.”

The company said it would ask its more than 2 billion users to rank the news organizations they trusted in order to prioritize “high-quality news” over less trusted sources. It said the new ranking system would seek to separate news organizations trusted only by their own subscribers from ones that are broadly trusted across society.

Facebook Chief Executive Mark Zuckerberg wrote in a blog post that the company was not “comfortable” deciding which news sources are the most trustworthy in a “world with so much division.”

“There’s too much sensationalism, misinformation and polarization in the world today,” he wrote.

“Social media enables people to spread information faster than ever before, and if we don’t specifically tackle these problems, then we end up amplifying them,” Zuckerberg added.

​Outside experts rejected

He said Facebook considered asking outside experts to choose the most reputable news sources, but that doing so would most likely have led to an “objectivity problem.” He said the company decided to rely on member surveys as the most “objective” way to rank trust in news sources.

Zuckerberg said it’s important that Facebook’s News Feed “promotes high-quality news that helps build a sense of common ground.”

He also announced that Facebook would shrink the content on its News Feed from 5 percent to 4 percent. This means users will see fewer posts from news organizations while scrolling through their feeds in favor of more posts from friends.

Facebook has been struggling with how to handle its distribution of news in an era of fake news and claims of media bias.

The social media company has faced accusations that it helped spread misinformation as well as Russian-linked content meant to influence the 2016 U.S. elections.

Also last year, U.S. Republican lawmakers expressed concern that Facebook was suppressing stories from conservative news sources.

The Pew Research Center has found that more than two-thirds of Americans are getting at least some of their news from social media, making such outlets prime sources of information.

Former IBM Developer Sentenced for Espionage, Theft of Trade Secrets

A former software engineer for IBM in China has been sentenced to five years in prison for stealing the source code for highly valuable software developed by the tech company, the U.S. Justice Department announced Friday.

Xu Jiaqiang, 31, was sentenced Thursday by a federal judge in White Plains, New York, months after he pleaded guilty to three counts of economic espionage and three counts of theft, possession and distribution of trade secrets.

Prosecutors said Xu stole the source code for computer performance-enhancing software while working for IBM from 2010 and 2014, with the intent to benefit China’s National Health and Family Planning Commission. 

Acting Assistant Attorney General Dana J. Boente of the Justice Department’s national security division said the agency “will not hesitate to pursue and prosecute those who steal from American businesses.”

Xu, a Chinese national, “is being held accountable for engaging in economic espionage against an American company,” Boente said in a statement.

U.S. Attorney Geoffrey S. Berman for the Southern District of New York said, “Xu’s prison sentence should be a red flag for anyone attempting to illegally peddle American expertise and intellectual property to foreign bidders.”

IBM was not identified in court documents. But a LinkedIn profile of Xu identifies him as a system developer for IBM in China from 2010 to 2014 with a master’s degree from the University of Delaware.

A Justice Department spokesman declined to say whether the company in question was IBM. IBM didn’t immediately respond to a request for comment.

Xu appeared on the FBI’s radar screen in 2014 after the bureau received a tip that Xu, who had by then left the company, claimed to have the source code to one of company’s most closely guarded software packages and was using it in “business ventures” unrelated to its clients.  

The software is described as a cluster file system sold to governments and large companies and used to enhance computer performance.

Undercover FBI agents posing as an investor and project manager for a large data storage company approached Xu, who tried to sell them the software and admitted that he’d built it with stolen source code, according to prosecutors.

IBM employees later confirmed to the FBI that the software had been built by someone with access to the company’s proprietary source code.

Xu was arrested in December 2015 after meeting with an undercover agent at a White Plains hotel.