Source: The Next Rembrandt/YouTube

From the Smithsonian comes news – and a must-see fascinating video – about a painting created using data from more than 168,000 fragments of Rembrandt’s work, trained to paint in Rembrandt’s signature style.

The video shows how the team used scanning, 3D printing, deep learning and inventiveness to create a “new” painting in Rembrandt’s style, and of a subject that Rembrandt would have chosen.

Sponsored by ING and Microsoft, with advice from TUDelft, Mauritshuis and the HET Rembrandthuis Museum, this video says it all.

Enjoy!

The use of Cortana, an AI personal assistant created by Microsoft initially for Windows phones, has been a big hit in Japan and China. The Mandarin language version, named Xiaoice, is being used daily by a TV station to report the morning weather.

Pronounced Shao-ice, meaning “little bing” in Mandarin, the youthful-sounding female voiced Xiaoice has been downloaded by over 20 million people and is being used as a chatbot throughout China on various platforms such as Weibo and 163.com, and now as a weather reporter for a Chinese TV station.

The NY Times said that “Microsoft has been able to give Xiaoice a more compelling personality and sense of intelligence by systematically mining the Chinese Internet for human conversations.” Microsoft said that the average user of Xiaoice talks to her more than 60 times a month!

In doing the weather report, Microsoft says Xiaoice calls upon smart cloud technologies and big data to learn about and interpret weather readings and then deliver reports with her “unique artificial intelligent style of emotional comments.” According to MS, the AI is more human-like than any other system of its kind, scoring 4.32 out of 5 in linguistic naturalness tests. For comparison, humans have an average score of 4.76.

MS's deep learning software enables Xiaoice to structuralize weather data and also interact with live scenes as weather news is visually broadcast (Xiaoice doesn't have a presence; no face or screen avatar; just a voice). Xiaoice also has an an emotional component that enables it to offer instant comments on weather data. The result is that Xiaoice has taken over the position of daily weather presenter on the morning Dragon TV “Morning News.”

“Xiaoice is a critical artificial intelligence product for Microsoft and she has been first launched in China and Japan.” Dr. Yongdong Wang said, “We feel thrilled to introduce Xiaoice into live TV and to jointly open a new chapter in the history of the scientific technology and television. We will pay more attention to practicality by developing artificial intelligence products available to everyone rather than just put forward mere sci-tech concepts. As Xiaoice is gradually penetrating into human life, engaging herself in more jobs and playing more social roles, Microsoft expects her to bring more bliss to human beings.”

It's a misnomer to call Xiaoice a robot although many in the media call it that. In reality, Xioaice is an AI software product.

“A robot requires physicality as a way to affect physical changes on its environment,” says GoogleX roboticist Travis Deyle. “An avatar or digital-only personality may possess artificial intelligence, but without a physical body it is not a robot. To put it another way: Robots often incorporate aspects of character development that are mainstays of digital avatars in much the same way that a robot incorporates machine learning, computer vision, mechanics, electronics, etc.  But we don't call every machine learning algorithm, computer vision system, mechanical apparatus, and printed circuit board a robot… as then the term “robot” loses all meaning.”

The MIT Technology Review’s annual list of 35 Innovators Under 35 aims to illustrate the most important emerging technologies of the moment. Released this August, the 2015 list features a number of robotics and AI visionaries. Check them out.

Lars Blackmore, Visionary, 34: SpaceX engineer, working on reusable rockets that could be flown back to Earth in reverse.

Yunji Chen, Inventor, 32: designing dedicated deep-learning processors, optimized “to compute the basic blocks of machine learning.”

Adam Coates, Visionary, 34: designing artificial intelligence for Chinese search company Baidu, to make the Internet more useful to millions.

Canan Dagdeviren, Inventor, 30: working on flexible sensors and batteries for a new class of medical device.

Travis Deyle, Inventor, 32: builder of robots that can be powered wirelessly. Now at Google X working on tech to improve health care.

You may remember our Robots Podcast interview with Travis about his research on Finding objects with RFID.

Rikky Muller, Entrepreneur, 34: designing the implantable hardware intended to interact directly with the brain and help treat mental disorders.

Jaime Shotton, Inventor, 34: computer visionary at Microsoft Research. Giving computers a new way to see the world.

Ilya Sutskever, Visionary, 29: pushing the frontiers of deep-learning for the Google Brain.

You may remember the Talking Machines Podcast interview with Ilya about Machine Learning and magical thinking.

Benjamin Tee, Inventor, 33: developing medical diagnostic and treatment technologies using a synthetic sense of touch.

Conor Walsh, Inventor, 33: working on robots that are soft, lightweight, and flexible so people can wear them to enhance their abilities.

You may remember our Robots Podcast interview with Conor’s colleague Donal Holland about their Soft Robotics Toolkit.

Melonee Wise, Entrepreneur, 33: Fetch Robotics CEO wants affordable robots for the warehouse and beyond.

Check out our coverage of Melonee and Fetch Robotics on Robohub.

Credit: Illustration by Magoz; Color portraits by Matthew Hollister; black-and-white portraits by Ping Zhu. Source: MIT Technology Review.

image: Thomas Hawk/flickr

In episode eighteen we talk with Sham Kakade, of Microsoft Research New England, about his expansive work that touches on everything from neuroscience to theoretical machine learning.

Ryan introduces us to active learning and we take a question on evolutionary algorithms.

https://soundcloud.com/talkingmachines/active-learning-and-machine-learning-in-neuroscience

Today we’re announcing that season two of Talking Machines is moving into development, but we need your help! 

In order to raise funds, we’ve opened the show up to sponsorship and started a Kickstarter and we’ve got some great nerd cred prizes to thank you with. But more than just getting you a totally sweet mug, your donation will fuel journalism about the reality of scientific research, something that is unfortunately hard to find. Lend a hand if you can!

image: Jer Thorp/flickr

In episode seventeen we talk with Jennifer Listgarten of  Microsoft Research New England about her work using machine learning to answer questions in biology.

Recently, she’s used machine learning to make CRISPR more efficient and sort populations in GWAS studies. We take a question from a listener about the development of computational biology and Ryan gives us some great advice on how to get into grad school (Spoiler alert: apply to the lab, not the program.)

https://soundcloud.com/talkingmachines/machine-learning-in-biology-and-getting-into-grad-school

Microsoft has been investing in Swiss research for years, and now the US technology company is renewing its longstanding collaboration with EPFL and ETH Zurich. Microsoft will provide one million Swiss francs per year in funding for IT-related research projects at the two universities over a period of five years. The collaboration is a continuation of a project launched in 2008 that focused on embedded technology software solutions and prototypes. The new collaboration will now broaden the areas of computer science and deepen the collaboration between the three partners.

Scientists in Lausanne and Zurich submitted 27 proposals, seven of which were selected by the steering committee. Among the successful projects:

Interacting with flying robots

Otmar Hilliges, a 34-year-old assistant professor, is one of the successful project applicants. The young computer scientist from ETH Zurich studies the interaction between people and computers. For his project, he and Dr. Shahram Izadi from Microsoft Research are investigating how flying robots can work and interact with people in active scenarios. Specifically, they want to develop a platform that enables flying robots to do more than just recognise people and navigate their way around them. Thanks to their algorithms, the robots should be able to react to gestures and touch as well.

Energy-efficient memory systems

At the EcoCloud center at EPFL, Dr. Edouard Bugnion and Professor Babak Falsafi are carrying out research into energy-efficient memory architectures for data centres that can handle huge amounts of data. To do this, they are combining thousands of energy-efficient micro-servers in a way that enables them to access the memories of the other servers with a minimal time delay. The two computer scientists are working with Dr. Dushyanth Narayanan and other scientists from Microsoft Research in Cambridge to develop new applications for this system, known as Scale-out NUMA.

They want to communicate more accurately, resolve issues faster and with higher levels of satisfaction than at present. They also want to enable more complex communication. And they want to do it amiably. As we move toward the Internet of Things, more players will enter the fray – all having the same goal.

Google appears to be acquiring people and companies to build a better human interaction interface. This interface can be the front-end of search, IoT devices, robotics and voice systems. Even if Google keeps their algorithms private, the first to deploy complex communication modules that can understand and resolve queries is going to have a big winner on their hands.

Certainly Google is going all out in this endeavor:

• Ray Kurzweil has been at work at Google for over a year now to develop natural language understanding. If successful, Google’s computers will be able to pick up semantic meaning enabling people to ask more complex questions.
• Just last week Google acquired DeepMind, a British AI start-up with some high-profile investors and a whiz kid founder (Demis Hassabis). DeepMind’s forte appears to be smart recommendations for online commerce.
 
• Facebook attempted to buy the deep learning group within DeepMind but Google outbid them.
• Deep learning is AI that trains AI to recognize patterns within streams of data instead of programming immense decision trees. For example, DeepMind trained software to play an Atari video game without teaching it the rules. The software had to learn through trial, errors and low scores.
• Google has hired and bought up leading researchers in AI for years including Sebastian Thrun, Peter Norvig and Geoffrey Hinton.

In a recent New Yorker article by Gary Marcus, Hyping Artificial Intelligence, Yet Again, it was clear that AI has received a lot of enthusiastic press attention over the years but hasn’t delivered upon those promises. Perhaps Google is on the right track toward converting all that desire into reality.

If you liked this article, you may also be interested in:

• Google acquires Nest for $3.2 billion
• Amazon isn’t the only delivery service testing flying robots
• Apple, Amazon and now Google: It’s a pretty exciting day for robotics!
• What were the top stories in robotics from 2013?
• Forget the octocopter, let’s talk about Kiva

See all the latest robotics news on Robohub, or sign up for our weekly newsletter.