The Best Artificial intelligence Up to Now

Oshua Bengio never sideline. As one of the three leading figures in the depth of learning the depth of the artificial intelligence, he has become a superstar to some extent. Artificial intelligence is an emerging field, and if you bring people from technology companies, multinational corporations and even the Department of Defense to a room, everyone will want to share their views on this topic.

But the other two leaders have long been engaged in the industry, Yann LeCun and Geoffrey Hinton had joined Facebook and Google, and now 53-year-old Yoshua Bengio chose to continue his third floor in the Montreal University campus teaching and learning. “I want to keep a neutral position,” Bengio said, drank a bit of licorice water, his desk filled with a wide range of documents, it seems chaotic.

Like the nuclear physicists of the last century, Bengio understood that the tools he was inventing were unparalleled in power and needed careful consideration and extensive consideration before it could be introduced. “I do not want to have one or two companies – here I will not name – the only big player in the transformation of artificial intelligence,” Bengio said, lifting the eyebrows, suggesting that we all understand what he said is the two companies The Left eyebrows said Menlo Parker (Facebook location), the right eyebrows that Mountain View (Google headquarters location). “This is no good for researching the community and is not good for humans.”

Bit coins purse company has financed 270 million

Bitchain wallet company Blockchain announced Thursday that it has raised $ 40 million ($ 270 million), venture capital firm Lakestar, Google Ventures and billionaire Richard Branson.

This investment coincided with the growing interest in encrypted currencies, especially bitcoin, whose prices have recently hit a record high since the beginning of the year.

Blockchain has developed a special currency wallet, which is actually a software that can store this digital currency and handle transactions between users.

According to Blockchain CEO Pitt Smith, $ 40 million in financing will be used to further expand the team and invest in R & D. In addition, the start-up company also hopes to open a new office in different countries to expand.

Smith said the recent rise in Bitcoin prices has caused demand for Blockchain wallet products to rise, and this round of financing will help the company meet these needs.

“We focus on expanding the scale to meet the market’s record needs,” Smith said to CNBC.

He said Blockchain has nearly 15 million registered users, “the monthly wallet activities worth billions of dollars”, including deposits and trading activities.

Smith also said the company’s total revenue this year is expected to grow by more than 1000%, but he did not disclose the specific figures.

Blockchain was founded in 2011 and now has 140 users worldwide. Its last round of financing in 2014, the scale of 30.5 million US dollars. Other investors in the latest round of financing include Nokota Management and Digital Money Group.

Waymo submits new evidence to the court

According to the United States famous technology media TechCrunch reported that Google (microblogging) Waymo recently submitted to the court a new document, claiming that before Uber acquired Otto, former Uber CEO Travis Kalanick (Travis Kalanick) Already know that joining Uber’s former Google engineer Anthony Levandowski (Anthony Levandowski) hands have Google’s technical secrets.

Legal action between Uber and Google Waymo continues. Waymo recently filed a document declaring that in March of last year, Travis Kalanik knew that Anthony Levandovsky had Google’s technical secrets.

Anthony Levandovsky created Otto, an unmanned truck start-up company. After Otto was acquired by Uber, Anthony Levandowski joined Uber and became head of the company’s unmanned department.

Waymo in this new document quotes Uber’s response to Waymo on June 8, which has the following words:

“On March 11, 2016, Levandovsky reported to Uber’s Kalanick, Nina Qi and Cameron Poetzscher and Lior Ron that he had confirmed that he had five CDs containing Google’s technical information. “

Google has introduced a new VR video format

When you look at 360-degree video, do you have to think about it when you do not want to look back or even look around? It does not matter, you are not a person, and now Google also thought of this problem, so they introduced a new video format “VR180”. From the name can be seen, the new format of the new video only half of the original 360-degree video perspective, users appreciate the content only need to pay attention to the immediate things.

Although the viewing angle is halved, the VR180 format video still needs to be recorded with two ultra-wide angle 180-degree cameras. The new format will support high resolution. VR180 by the Google Tray VR team and Daydream team jointly launched, Daydream is with Lenovo, LG and small ants cooperation, the corresponding camera products will be listed this winter, Google will also open to third-party manufacturers equipment certification program.

Speaking of VR180 format itself, in the general plane screen, the user can see the 180-degree video center position, and the viewing angle will not change, and if you wear VR glasses, the user can turn around to see the center position outside the screen, The video boundary is displayed in black.

LG is expected to achieve better robot navigation and VR tracking

Augmented Pixels invented a new technology that uses computer vision navigation, and LG Electronics announced that it has developed a 3D camera module for assisting robots based on this technology. Augmented Pixels is a computer vision research and development company headquartered in Palo Alto, California, which calls this new technology SLAM technology, the so-called SLAM technology, is based on camera and laser radar two visual technology for 3D modeling, to help objects In the unknown environment inside the positioning and tracking, to avoid drift, which is the key to intelligent machine walking movement.

The company is working on the technology used in robots, unmanned aerial vehicles, and AR and VR equipment. The 3D module can also be used for AR and VR head and outside the track, which means that the head will be able to use the camera to identify their own location. LG has developed a compact module consisting of stereo cameras, IR and airborne processors designed to take into account both low power and superior performance. In addition, it can be customized for different hardware platforms and use cases. Augmented Pixels provides software for autonomous navigation (obstacle avoidance, semantic recognition of point cloud, etc.) based on its specialized SLAM technology.

In order to strengthen its virtual reality products, Adobe from the VR software company Mettle acquired SkyBox technology. Mettle 2015 launches SkyBox plug-ins for Adobe Premiere Pro CC and Adobe After Effects CC post-production and complements Adobe Creative Cloud’s existing 360 / VR movie production technology. The technology is also used in the New York Times, CNN, HBO and YouTube. As part of the terms of the acquisition, Mettle co-founder Chris Bobotis will join Adobe as a professional video director. “This is an important step in our partnership,” Bobotis said in a statement. “Adobe has been strengthening Creative Cloud’s VR and its applications in recent months. In November, it announced several updates, including a new 3D design application update and three new Android applications, as well as a series of desktop and mobile device updates. “We believe that virtual reality content should be as simple as possible for creators,” said Steven Warner, vice president of digital video at Adobe. “The acquisition of Mettle SkyBox technology enables us to provide a more integrated VR editing and performance experience for the film and video community.”

Interpreting Amazon and Google’s Price War

Finally, Amazon and Google in the smart home market price war broke out, has been high on the Echo and Google Home have been approaching the $ 100 price level. However, from China’s IoT first platform – graffiti intelligence said that Chinese enterprises may become the biggest beneficiaries.

In early June, Apple WWDC2017 announced the price of $ 349 HomePod will be listed at the end of the year, it was speculated that the market at the end of the outbreak of the price war, did not expect the New Year, the price war on the fight. First Amazon will be the original price of 179 US dollars Echo fell to 139 dollars, while expanding the price gap with Apple HomePod at the same time, approaching Google’s Google Home; but Google is not simple, Google home quickly adjusted to 109 dollars, the decline has reached 20 US dollars, while approaching the critical line of $ 100.

In this way, Apple’s HomePod is very embarrassing, and do not say the price of 349 US dollars to Echo 2.5 times, Google Home 3 times, from the end of December in the three major markets listed for half a year, then if Amazon and Google will be cruel to its products down to 100 dollars, the consumer’s choice will be significantly inclined to the latter two.

Amazon and Google are attracting top academic scientists

Nathan Price, now deputy director of ISB, which works for Providence Medical Services, says four out of a group of researchers have recently left Amazon, Google and Microsoft.

Price has found a way to follow the trend, rather than to the Institute of outstanding biologists and bioinformatics title and struggle.

“We are asking talented researchers to spend a few years with us and develop some cool things before they are hired,” he said in an interview.

Technology companies are demanding, these companies are competing to cloud computing and machine learning to 3 trillion US dollars in the medical market. For example, ISB uses Amazon Web services and Google’s cloud technology on a variety of projects.

“As technology companies sign contracts with large health systems to manage the growing and most sensitive data, they need employees to understand specific topics such as protein networks and genetic networks,” he said.

In the ISB, co-founded by genetics pioneer Leroy “Lee” Hood, researchers have gained engineering knowledge through the acquisition of engineering courses from educational websites such as Coursera to improve their scientific knowledge.

Google added job function in its search engine

New features appear on desktop and mobile search, only in English. Use the new features to enter only information, such as “near my job”, “writing position” or similar information, the search results page will display a new job search plugin, allowing you to view a lot of job opportunities. You can further refine the requirements, for example, you can only show full-time work. When you click to get specific information about a job, you can see Glassdoor or Indeed rating the relevant company.

You can also according to the industry, location, information release time, the employer to filter the job. If you feel that the search request is just right, you can open the “notification” function, when the new job in line with your personal requirements, the system will automatically notify you.

Nick Zakrasek, Google product manager for the project, said: “Looking for a job like a date, everyone has a unique request that only one person can get the job.”

In order to create a comprehensive list, Google first of all employers in the major job sites posted duplicate information removed, and then machine learning training algorithm to view information, to information classification. Job search sites often have some information to mark, to help the search engine to understand, let it know what the content is recruitment information.

Google can handle a variety of tasks for voice and images

Recently, Google released a low-key academic papers, in the paper depicts the machine learning blueprint. Google has developed a new machine learning system, which is called “a multifunctional model for learning all tasks.” This model provides a template for later research on how to create a machine that can handle multiple tasks well model.
As Google researchers call, the “multifunctional model” accepts a variety of task training, including translation, language analysis, speech recognition, image recognition and target detection. While its results do not show a fundamental improvement in the existing approach, it is at least that training the machine learning system on different tasks helps to improve its overall performance.
The accuracy of the “multifunctional model” in machine translation, speech and grammar analysis has been improved compared to training on a single computing machine.
Google’s paper can provide a case for the future development of machine learning systems that can be applied more widely and may be more accurate than most of the narrow solutions on the market today. More importantly, these techniques (or their derivatives) can help reduce the amount of training data required to train a viable machine learning algorithm.