China's Race for Artificial Intelligence (AI) Technology

[TaiShang]

Baidu gets ex-Google scientist Ng in race for AI research

Chinese search giant Baidu - often called the Google of China - has doubled down in the Artificial Intelligence (AI) race. Not only have they hired away Google's top scientist Andrew Ng in the field, but they are starting a new research center in Google's Silicon Valley backyard.

Baidu will put $300 million into the project and Ng, and his long-time collaborator Adam Coats, are at work building a new staff there. "I feel excited about the work and I'm honored Adam has joined me," Ng said in a telephone interview with China Daily.

Ng's specialty is an esoteric field that embraces several abstract names: machine learning, deep learning, unsupervised learning, autonomous artificial intelligence, that basically mean trying to get computers to teach themselves without being specifically programmed. The holy grail of the field is an algorithm that will mimic how the brain works.

"Most of us use algorithms dozens of times a day without realizing it," Ng said. "Every time you check your email and your spam folder saves you from going through hundreds of spam emails, every time your cell phone camera auto-focuses on your friend's face, that's machine learning."

Deep learning is the technology that specifically takes inspiration from how the brain's neural network operates, the goal being to "build software that learns from data," Ng said. "And in the past few years, I think deep learning has created substantial economic value.

"At web search engines like Baidu it is allowing us to surf more website pages and force better ads to users. It's a technology that is taking the machine learning world by storm, and it's something we plan to build on here in our research," Ng said.

Ng first got interested in AI as a 16-year-old intern for a professor at the National University of Singapore, where he helped implement a neural network. "Since then," he said, "I just thought what more meaningful thing could there be for me to work on than to make computers smarter so that they can help people more."

As a professor at Stanford University, Ng's access to computers limited his research to relatively small neural networks. So he looked around Silicon Valley.

"It turned out Google had a lot of computers," he said. "So I started a project to build a much larger deep learning facility, 100 times bigger than what academia previously had been able to do. "

That project, started in 2011, was dubbed "Google Brain" and one of the team's early successes came when it connected 16,000 computer processors into a "neural network" model of a brain - at the time the largest of its kind in the world.

"Imagine, if you will, it's like a little simulated baby brain and it wakes up not knowing anything and what we decided to do was make it watch YouTube for a week and after a week we would probe it to try and figure out what it had learned," Ng said.

The team expected the network to begin to recognize the most common image on YouTube - the human face - which it did, but to their complete surprise, the machine taught itself to also recognize cats.

"The remarkable thing about this was that no one had ever told it what a cat is," Ng said. "It had discovered the concept of a cat by itself."

Deep Learning is broadly broken into two categories, Ng explained. Learning from "tagged data", presorted and labeled units of information like say 50,000 pictures of cars. The state-of-the-art software out there for soaking up huge amounts of tagged data, Ng said, is pretty good.

"This is what has driven the performance improvement in speech recognition, in web search and in identifying the most relevant ads," Ng said, noting that Baidu has a large operation in Beijing that has been doing groundbreaking work in learning from tagged data for years.

"Those investments have paid off well over the years and we'll certainly continue to invest in that," he said.

Ng says the focus of the new Sunnyvale center will be more on learning from "untagged data", which is modeled after the way humans actually learn. "Instead of having a parent point out every object to you every moment of the day, most of what you learn is from going out and seeing and experiencing the world for yourself," Ng said.

"It is closer to how we believe humans and animals learn, and so it has more of a potential for larger breakthroughs in AI," he said.

 

[TaiShang]

Chinese Search Company Baidu Built a Giant Artificial-Intelligence Supercomputer | MIT Technology Review

Chinese search giant Baidu says it has invented a powerful supercomputer that brings new muscle to an artificial-intelligence technique giving software more power to understand speech, images, and written language.

The new computer, called Minwa and located in Beijing, has 72 powerful processors and 144 graphics processors, known as GPUs. Late Monday, Baidu released a paper claiming that the computer had been used to train machine-learning software that set a new record for recognizing images, beating a previous mark set by Google.

“Our company is now leading the race in computer intelligence,” said Ren Wu, a Baidu scientist working on the project, speaking at the Embedded Vision Summit on Tuesday. Minwa’s computational power would probably put it among the 300 most powerful computers in the world if it weren’t specialized for deep learning, said Wu. “I think this is the fastest supercomputer dedicated to deep learning,” he said. “We have great power in our hands—much greater than our competitors.”

Computing power matters in the world of deep learning, which has produced breakthroughs in speech, image, and face recognition and improved the image-search and speech-recognition services offered by Google and Baidu.

The technique is a souped-up version of an approach first established decades ago, in which data is processed by a network of artificial neurons that manage information in ways loosely inspired by biological brains. Deep learning involves using larger neural networks than before, arranged in hierarchical layers, and training them with significantly larger collections of data, such as photos, text documents, or recorded speech.

So far, bigger data sets and networks appear to always be better for this technology, said Wu. That’s one way it differs from previous machine-learning techniques, which had begun to produce diminishing returns with larger data sets. “Once you scaled your data beyond a certain point, you couldn’t see any improvement,” said Wu. “With deep learning, it just keeps going up.” Baidu says that Minwa makes it practical to create an artificial neural network with hundreds of billions of connections—hundreds of times more than any network built before.

A paper released Monday is intended to provide a taste of what Minwa’s extra oomph can do. It describes how the supercomputer was used to train a neural network that set a new record on a standard benchmark for image-recognition software. The ImageNet Classification Challenge, as it is called, involves training software on a collection of 1.5 million labeled images in 1,000 different categories, and then asking that software to use what it learned to label 100,000 images it has not seen before.

Software is compared on the basis of how often its top five guesses for a given image miss the correct answer. The system trained on Baidu’s new computer was wrong only 4.58 percent of the time. The previous best was 4.82 percent, reported by Google in March. One month before that, Microsoft had reported achieving 4.94 percent, becoming the first to better average human performance of 5.1 percent.

Wu said that as well as thinking about how to make Minwa even larger and use it on video and text, Baidu’s researchers are working on ways to shrink their trained neural networks so they can operate on mobile devices.

He showed a video of a prototype smartphone app that can recognize different breeds of dog, using a condensed version of a deep-learning network trained on a predecessor to Minwa. “If you know how to tap the computational power of a phone’s GPUs, you can actually recognize on the fly directly from the image sensor,” he said.

 

[Raphael]

Chinese search big Baidu unveils advanced AI - NY Daily News

Chinese web search giant Baidu unveiled its latest technology Monday, saying it had taken the lead in the global race for true artificial intelligence.

Minwa, the company's supercomputer, scanned more than 1 million images and taught itself to sort them into about 1,000 categories — and did so with 95.42% accuracy, the company claims, adding that no other computer has completed the task at that same level.

Google's system scored a 95.2% and Microsoft's, a 95.06%, Baidu said.

All three companies' computers, however, exceed human performance.

The concept of "deep learning," or self-learning, algorithms is not unique to Minwa. Yet Baidu seems to have the upper hand and is not slowing down: the company has announced plans to build an even faster computer in the next 2 years, one capable of 7 quadrillion calculations per second.

 

[cirr]

Baidu’s Artificial-Intelligence Supercomputer Beats Google at Image Recognition

A supercomputer specialized for the machine-learning technique known as deep learning could help software understand us better.

By Tom Simonite on May 13, 2015

Chinese search company Baidu built this computer to accelerate its artificial-intelligence research.


Chinese search giant Baidu says it has invented a powerful supercomputer that brings new muscle to an artificial-intelligence technique giving software more power to understand speech, images, and written language.

The new computer, called Minwa and located in Beijing, has 72 powerful processors and 144 graphics processors, known as GPUs. Late Monday, Baidu released a paper claiming that the computer had been used to train machine-learning software that set a new record for recognizing images, beating a previous mark set by Google.

“Our company is now leading the race in computer intelligence,” said Ren Wu, a Baidu scientist working on the project, speaking at the Embedded Vision Summit on Tuesday. Minwa’s computational power would probably put it among the 300 most powerful computers in the world if it weren’t specialized for deep learning, said Wu. “I think this is the fastest supercomputer dedicated to deep learning,” he said. “We have great power in our hands—much greater than our competitors.”

Computing power matters in the world of deep learning, which has produced breakthroughs in speech, image, and face recognition and improved the image-search and speech-recognition services offered by Google and Baidu.

The technique is a souped-up version of an approach first established decades ago, in which data is processed by a network of artificial neurons that manage information in ways loosely inspired by biological brains. Deep learning involves using larger neural networks than before, arranged in hierarchical layers, and training them with significantly larger collections of data, such as photos, text documents, or recorded speech.

So far, bigger data sets and networks appear to always be better for this technology, said Wu. That’s one way it differs from previous machine-learning techniques, which had begun to produce diminishing returns with larger data sets. “Once you scaled your data beyond a certain point, you couldn’t see any improvement,” said Wu. “With deep learning, it just keeps going up.” Baidu says that Minwa makes it practical to create an artificial neural network with hundreds of billions of connections—hundreds of times more than any network built before.

A paper released Monday is intended to provide a taste of what Minwa’s extra oomph can do. It describes how the supercomputer was used to train a neural network that set a new record on a standard benchmark for image-recognition software. The ImageNet Classification Challenge, as it is called, involves training software on a collection of 1.5 million labeled images in 1,000 different categories, and then asking that software to use what it learned to label 100,000 images it has not seen before.

Software is compared on the basis of how often its top five guesses for a given image miss the correct answer. The system trained on Baidu’s new computer was wrong only 4.58 percent of the time. The previous best was 4.82 percent,reported by Google in March. One month before that, Microsoft had reportedachieving 4.94 percent, becoming the first to better average human performance of 5.1 percent.

Wu said that Minwa had made it possible to train the system on higher-resolution images. It also permitted use of a technique that turned the original 1.2 million training images into two billion by distorting them, flipping them, and altering their colors. Using that larger training set improved accuracy by preventing the system from becoming too fixated on the exact details of the training images, said Wu. The resulting system should be better at handling real-world photos, he said.

As those slim margins of victory on the ImageNet challenge might suggest, deep learning is now ready for tougher challenges than image recognition, such as interpreting video or describing images in sentences (see “Google’s Brain-Inspired Software Describes What It Sees in Complex Images”). Wu said that as well as thinking about how to make Minwa even larger and use it on video and text, Baidu’s researchers are working on ways to shrink their trained neural networks so they can operate on mobile devices.

He showed a video of a prototype smartphone app that can recognize different breeds of dog, using a condensed version of a deep-learning network trained on a predecessor to Minwa. “If you know how to tap the computational power of a phone’s GPUs, you can actually recognize on the fly directly from the image sensor,” he said.

Chinese Search Company Baidu Built a Giant Artificial-Intelligence Supercomputer | MIT Technology Review

 

[Economic superpower]

Well done to Baidu. Shows China's technological progression.

 

[大汉奸柳传志]

Never liked Baidu，I use Bing for search

Theyhave great potentials，if put in capable hands

 

[cirr]

Perhaps it would be to our advantage if we simply send all rubbish the West's way。

 

[qwerrty]

he's a supa genius artist, the modern day picasso. they should give him citizenship and nobel prize

 

[Steve781]

Ai Weiwei denied six-month UK visa - BBC News

Dissident Chinese artist Ai Weiwei has been refused a six-month visa by the British government amid claims he lied on his visa application form.

Ai posted a letter on Instagram stating his entry to the UK had been restricted because he failed to declare his "criminal conviction".

But Ai, though detained in China for 81 days in 2011, has never been charged or convicted of a crime in China.

The artist was granted a 20-day visa to attend the opening of his London show.

However, it may mean he cannot supervise the installation of the landmark solo exhibition at the Royal Academy of Arts which has a private preview on 15 September.

The letter, from the visa section of the British Embassy in Beijing, stated Ai's entry to the UK "has been restricted to the requested dates of travel... because you have failed to meet the business visitor rules".

The document, signed by an entry clearance manager from the UK Visas and Immigration department, continued: "It is a matter of public record that you have previously received a criminal conviction in China, and you have not declared this.

"While an exception has been made in this instance, any future application you submit must be completed as accurately as possible," the letter concluded, adding the artist might otherwise face a 10-year ban if he did not comply.

In a separate post on Instagram, Ai stated he "has never been charged or convicted of a crime" and had "attempted to clarify this claim with the UK Visas and Immigration Department and the British Embassy in Beijing over several telephone conversations".

"But the representatives insisted on the accuracy of their sources and refused to admit any misjudgement. This decision is a denial of Ai Weiwei's rights as an ordinary citizen," he added.

Last week, Ai had his passport returned to him after it was confiscated by authorities four years ago.

It was taken when he was arrested in 2011 during a government crackdown on political activists. He was held over alleged crimes of bigamy and tax evasion, but was released without charge.

Ai was fined 15m yuan ($2.4m, £1.55m) for tax evasion in a civil case in 2012. The artist lost an appeal against the fine - which he maintains was politically motivated in retaliation for his criticism of the Chinese government.

In a statement, Britain's Home Office said visa applications were considered "on their individual merits and in line with the relevant legislation".

It added: "Mr Ai has been granted a visa for the full duration of his requested dates of travel".

Ai's 20-day visa means he will not be in the UK when China's President Xi Jinping make a high-profile state visit in October - potentially avoiding any diplomatic embarrassment at a time when Britain is trying to improve relations with China.

The BBC has contacted the British Embassy in Beijing to query its decision, but has yet to receive a response.


----------------------------------------------
Lately the UK has tried very hard to please us: joining the AIIB, refusing to meddle in HK riots, shunning the Delhi Llama, and finally rejecting AWW's visa application.

What they don't realize is that we have no desire to restrict AWW's travel. Actually, we gave him a passport because we want him to GTFO of our soil as soon as possible, and Britain is as good a place to exile him as anywhere else.

China also recently refused to sell Argentina anti ship missiles which was very surprising to me.

 

[Jlaw]

he's a supa genius artist, the modern day picasso. they should give citizenship and nobel prize

sure if you call taking pictures with your middle finger with buildings in the background a work of genius

 

[qwerrty]

sure if you call taking pictures with your middle finger with buildings in the background a work of genius

and naked

 

[Raphael]

China also recently refused to sell Argentina anti ship missiles which was very surprising to me.

It's a quid pro quo.

 

[sicsheep]

The UK has done something very odd: It denied a business visa to renowned Chinese dissident artist Ai Weiwei.

Ai, on Wednesday, posted the letter from the UK foreign office denying the visa on his Instagram account. It informs him that his application for a six-month business visa has been denied because he lied on his visa application when he said he had not been charged or convicted of any crime. Instead, the letter says, he'll be granted a temporary visa for a few weeks in September:

This is really strange.

The letter says that Ai's application was denied because it is a "matter of public record" that the artist had received a criminal conviction in China, and that he had failed to disclose that on his visa form. But not only is that not a matter of public record, it doesn't appear to be true at all. Although the artist has been fined, imprisoned, and persecuted by Chinese authorities, all for his artistic and political activity, he has never actually been convicted of a crime.

"Under Chinese law Ai’s case ended in the police investigation stage and has not reached the court," Chinese human rights lawyer Liu Xiaoyuan explained to the Guardian. "The case does not have a court sentence and hence by Chinese standard, Ai doesn’t have a criminal conviction."

Indeed, it doesn't appear that Ai has ever even been formally charged with a crime, much less convicted of one. In 2011, he was arrested and detained for 81 days. He was investigated and held on suspicion of a variety of different charges including tax evasion, but was never formally charged and was eventually released.

But more importantly, it is also a matter of "public record" that the investigation was an act of political persecution meant to pressure one of China's highest-profile dissidents into silence.

The UK immigration authorities clearly knew that, because the letter makes a big deal out of the decision to "exceptionally" grant entry clearance "outside the immigration rules" — whereas the normal response, as the letter snidely points out, is to reject the application entirely and ban the applicant from reapplying for a decade.

In other words, this isn't an official being an unreasonable stickler for the letter of the law. It's an official making up a supposed violation of the law, then bending over backward to offer an extralegal solution to that imagined problem.

Perhaps this was just a mistake. Maybe the official who processed the visa application erroneously thought that Ai had been convicted and never bothered to get the facts straight. But it would have been easy to ask for clarification, or more evidence, rather than simply rejecting the application. Surely such a high-profile visa request is an occasion for double-checking one's facts — or at least making a phone call?

Or perhaps something else is going on. One possible factor in all this is that Chinese president Xi Jinping is due to make a state visit to London in October. If the UK issued Ai a six-month business visa, that would have allowed him to stay in the country through Xi's visit. As the Financial Times reports, "Any protest or public stunt by the artist could have caused embarrassment to the British government, which has worked hard to improve relations with Beijing in recent years."

This is baffling: the UK just denied a visa to Chinese dissident artist Ai Weiwei - Vox

@mike2000 is back

 

[21stCentury]

Is that the guy who smashed cultural relics i.e. Ming dynasty vases? for "artistic expression"??

 

[opruh]

Britain is a good pet it seems. @mike2000 is back