Machine learning
April 30, 2017
There is something uncanny about the way this dog-like robot moves – its skeletal frame whirs loudly as it marches on the spot, then moves side to side, and around in a circle in a strange dance. Built by a team in the Robotic Systems Lab at the Swiss Federal Institute of Technology (ETH Zurich), assistant professor Marco Hutter says the “ANYmal” is his newest creation.
Not only can it run but climb, crouch and jump. “We wanted to make something that was optimal from a robotics point of view,” he says. “We put springs in all the joints so we can use it in all sorts of environments.” As part of a pilot project, the ANYmal has been put to work on offshore oil and gas platforms where it can go about inspection tasks (often dangerous for humans) completely autonomously thanks to laser sensors and cameras.
I ask how it compares with the robot that was sent to Mars. “In general, space technology is very old,” says Hutter, walking me down the corridor and pointing to a dusty old unit on caterpillar tracks. “This was part of a study we were doing for the European Space Agency. But wheels are boring – legs are the future.”
Intelligence Hub
Founded in 1855, the ETH is Switzerland’s answer to MIT. Ranked one of the best universities in the world, more than 20 Nobel Prizes have been awarded to its alumni over the years, including Albert Einstein in 1921. Today it has 20,000 students and an annual budget of Sfr 1.7 billion ($1.7 billion), funded by taxpayers. “That is part of the reason the ETH is the best,” says professor Peter Seitz, a “sherpa” from its Innovation and Entrepreneurship Lab (ieLab).
In a warehouse on the Science City campus, a short drive northwest of the old town, architects are using giant mechanical arms to explore new construction techniques that employ nothing more than loops of yarn and pebbles, for example, or 3D printed concrete. Aleksandra Anna Apolinarska, an architect in the Gramazio Kohler Research Lab at the ETH Zurich, says the days of mass production are behind us. “We think it is time for mass customization.”
From self-driving cars to augmented reality, the ETH is forging a new tomorrow in myriad ways. And with the help of ieLab, Seitz’s students have the opportunity to take ideas from the research stage to market. Between 1996 and 2016, 355 spin-off companies have been founded at the ETH, a number of which have been in the field of robotics.
It’s no wonder that Chris Anderson, CEO of 3D Robotics and former editor-in-chief of Wired, has dubbed Zurich “the Silicon Valley of Robotics.”
Talent Contest
In 2016, Switzerland was ranked first in Cornell University’s Global Innovation Index, and Zurich came second in the Mercer Quality of Living survey, significantly ahead of San Francisco (28th position).
Unsurprisingly, over the decades, the ETH has provided a compelling reason for big companies to locate themselves in pretty little Zurich, a city of just 380,000 people that has grown into an international hub for banking, finance and innovation. The renowned IBM Research Lab was the company’s first outside the US when it opened here in 1956.
Chris Sciacca, IBM Research’s communications manager for EMEA, says, “We chose Switzerland because of the access to talent and skills that the country affords us. The standard of living is very good and the government is fantastic at supporting science and innovation with grants. It is very stable, democratic and open. All this means you can attract the best and the brightest.”
From his pocket, Sciacca pulls a gold medallion. It’s one of IBM Zurich’s four Nobel Prizes, two of which were won in the mid 1980s for the invention of high temperature superconductivity and the nanoscale microscope. “Up until this you really couldn’t see atoms and molecules with good resolution. You can really point to the 30-year history of nanotechnology in Switzerland to this invention,” he says.
Independent Thinking
The level of innovation going on at IBM is mind-blowing – in 2016, its inventors were awarded a record 8,088 patents in the US alone, more than any other company (Samsung was in second place with 5,518 and Canon third with 3,665). Interestingly, more than 2,700 patents were related to artificial intelligence, machine learning and cloud computing. In between forkfuls of risotto, Alessandro Curioni, IBM fellow, vice-president Europe and director of IBM Research Zurich, gives me a crash course in cognitive computing.
“The way we interact with computers is changing,” he explains. “First it was tabular computing, then the programming era, now it is natural language. The ability to analyze unstructured data [such as images and sounds] will accelerate by an order of magnitude the research and development in every field, including aviation and space travel.”
The new Cognitive era began in 2011, when IBM’s Watson super-computer won the TV quiz show Jeopardy. Eleni Pratsini, director of cognitive IoT solutions at IBM Research, says: “One of the rules of the game was that Watson was not connected to the Internet so scientists had to feed it hundreds of books and teach it to reason like a human, to understand riddles, puns and subtle connotations.”
The breadth of this human-like AI, which can make associations and learn, has since been expanded – now you can logon to ibm.com/watson/developercloud and access more than 60 versions of Watson in the form of individual APIs created for specific tasks such as image recognition and personality insight. Want to build a chat bot? Download the Conversation API and get to work. Hilton is already using AI to power Connie, its robotic concierge in Virginia.
Tomorrow’s World
Come Saturday, I take a trip to the public Thermalbad and Spa. Down in the vaults of this former brewery, locals soak in giant repurposed wooden vats. At the same time, in one of the buildings across the way, a solo employee pounds away in a gym at the otherwise peaceful Google campus.
Since 2004, Zurich has been the home of Google’s largest engineering base outside the US (the biggest is Mountain View in California and the second-largest New York City). Engineering director Emmanuel Mogenet heads up the company’s new European Research Lab, which was set up last year on the existing office campus.
Operating in parallel to IBM (not collaborating but not competing), Google has chosen the Swiss city to host its first lab outside the US dedicated to AI, computer perception and machine learning (with the exception of Deep Mind in London, an AI start-up that was acquired by Google in 2014). Why? Because the ETH “produces the best computer scientists in Europe,” says Mogenet.
To make sure they not only attract but retain them, the company goes out of its way to provide not just gyms and free food, but fantasy work environments complete with fireman’s poles and slides, and egg-shaped privacy pods. “Our basic philosophy is that you are most productive when you are enjoying yourself,” Mogenet says. “It is extremely informal – there are a lot of people who wear slippers at the office and bring their dog in.”
At the moment, there are 2,000 people representing 75 nationalities working here, but this number is set to rise to 5,000 “Zooglers” with the opening of its new offices in Europaallee, by Zurich Hauptbahnhof station. Andreas Meyer, CEO of Swiss Federal Railways, says: “The district around the main station in Zurich will be a hotspot where innovative services are developed and tested, and the future is significantly shaped.”
For example, nearby is the Technopark, a half-million-square-foot site that is home to 300 start-ups all hoping to become a success story. Last year, Facebook bought local computer vision venture Zurich Eye, which was founded by three members of the University of Zurich’s Robotics and Perception group. Although the social network has its main Swiss office in Geneva, it is opening a small base for its Oculus virtual reality subsidiary here. If you’re smart, you’ll get in on the action too.
By Jenny Southan