Technology Quarterly | Brain scan

Teaching tomorrow

Sebastian Thrun: The pioneer of Google’s autonomous cars wants to teach people how to face the future

“BECAUSE of the increased efficiency of machines, it is getting harder and harder for a human to make a productive contribution to society,” says Sebastian Thrun. This is what you might expect to hear from the man who suggested Google’s controversial Street View project to photograph the world’s roadsides, who developed the company’s eerie self-driving cars and who founded the secretive Google “skunk-works” project responsible for Glass, a wearable computer that resembles spectacles. Yet that does not mean Mr Thrun is in thrall to the march of the machines. “To the extent we are seeing the beginning of a battle between artificial intelligence (AI) and humanity, I am 100% loyal to people,” he says.

Long before he felt compelled to pledge his allegiance to the human race, Mr Thrun was fascinated by the interaction of people and robots. Mr Thrun was born in Germany in 1967 and as a graduate student at the University of Bonn designed RHINO, a cleaning robot that could explore an office it had never been in before, build a digital map of its floors and locate items such as drinks cans or balls of paper. Lacking arms, it simply barked out its intention to clean up the mess—or perhaps to encourage humans nearby to muck in.

In 1994 Mr Thrun entered RHINO into a competition organised by the Association for the Advancement of Artificial Intelligence, in Seattle. The robot was a success, not only helping tidy up the contest’s intended conference room but venturing out into hallways to find more rubbish. RHINO was succeeded by a brace of experimental robotic guides, complete with voices and faces that smiled and frowned, to interpret exhibits for museum visitors in Bonn and Washington, DC.

These innovations attracted the attention of Stanford University in California, which in 2004 invited Mr Thrun to become the head of its revitalised Stanford Artificial Intelligence Laboratory (SAIL). Given a throng of enthusiastic students and free rein to “change the way we understand the world”, Mr Thrun chose to focus on a follow up to the Grand Challenge, an off-road race for self-driving vehicles organised by DARPA, the Pentagon’s R&D wing.

“At the time, there were a few, little-noticed people who had this vision to transform the world by making cars drive themselves,” says Mr Thrun. They consisted mostly of university researchers and enthusiasts with little funding and even less exposure. Their cars were equally uninspiring. One flipped itself over before the first DARPA race in 2004 started and soon others broke down or became stuck. Even the best completed just seven miles of the 150-mile course.

The follow-up event in 2005 was a different story. Mr Thrun’s SAIL group modified a Volkswagen sports-utility vehicle with lidar, which works like radar but uses light waves to bounce off features and thereby build up a much more detailed picture of its surroundings. The car, nicknamed Stanley, also had video cameras, wheel sensors and GPS navigation. It crossed the finish line of a (slightly easier) 132-mile course first, in under seven hours. It was also during this desert race that Mr Thrun met and befriended Larry Page, co-founder of Google.

Stanley and its rivals proved so competent that in 2007 DARPA held a third contest for autonomous cars, the Urban Challenge. This required vehicles to navigate 60 miles of city streets, avoiding traffic and obeying all the rules of the road. This time around, SAIL’s car, a Volkswagen estate dubbed Junior, took only second place. But in developing it, Mr Thrun discovered that car-mounted cameras pointing in all directions enabled his team more easily to understand and debug Junior’s performance. “And then at some point we realised the images alone were really interesting, that they might allow you to travel to distant places,” says Mr Thrun.

Picture this

Mr Page agreed, and lured Mr Thrun out of academia to work on Google’s project to photograph the world. Google Street View quickly became the biggest image database ever built. In the process, it captured more than just landmarks and shopfronts; it immortalised individuals leaving strip clubs, political protesters, sunbathers and people being arrested. Street View vehicles also illegally (but inadvertently, says Google) collected Wi-Fi data transmissions from millions of private users. The company subsequently faced legal and regulatory action around the world.

By this time, Mr Thrun had moved on. “The obvious next thing was to use these methods for something other than just sightseeing and helping people find their way around,” he says. Google had decided to build its own self-driving car. Mr Thrun assembled a team of engineers from Silicon Valley, and Google purchased a small startup from Berkeley, called 510 Systems, that had demonstrated a self-driving Toyota Prius. This formed the nucleus of a fleet of autonomous Prius hybrids that were soon criss-crossing the Bay Area, using Mr Thrun’s mapping technology.

Imagine how many weeks of your life you waste being in traffic jams

Unlike traditional carmakers, which were inching their way towards autonomy with collision avoidance and lane-keeping systems, Mr Thrun’s vision was always a fully self-driving car. “Imagine how many weeks of your life you waste being in traffic jams, that you could use productively, or watch a movie or sleep,” he says. “Then think, if a car could come to you empty, would you still own one? We could live in a much better society if there was less personal car ownership.”

Mr Thrun points out that cars are typically being driven on the road for just a fraction of time, and that even motorways that appear jam-packed are only 5-10% physically occupied. Intelligent, self-driving cars could unlock existing capacity in cities, meaning less road-building and fewer cars and car parks. But the great promise of self-driving cars is their supposed ability to reduce accidents that now kill over 1m people globally each year. “At this point, I can happily and proudly say that Google’s cars are clearly better drivers than me,” says Mr Thrun.

With his self-driving cars making rapid progress, Mr Thrun was given responsibility for nurturing other innovative concepts at a secretive new division of Google that came to be known as Google [x]. Google [x] projects that have been publicly announced include airborne wind turbines, diagnostic contact lenses, wireless internet for remote areas using high-altitude balloons and surgical robotics. “We ran a whole bunch of experiments, crazy moonshots that rational people would probably agree could never be done, just to see how far we could get,” he says. “While not every project has been a success [and] there have been many failures, I still believe we made substantial progress.”

Mr Thrun says that the biggest difference between Google[x] and a research institute is Google’s drive to take ideas from the drawing board to a finished product. The only example, so far, is Google Glass—lauded and derided in equal measure for its miniaturised heads-up display, speech control and pervasive video capture. Despite the public’s tepid reaction to Glass, Mr Thrun believes that prosthetic devices capable of recording, digitising and transcribing a person’s entire life are already technically feasible.

Recently, however, Mr Thrun has come to realise that such things come at a price. “Machines will eventually outsmart people in every dimension. They are becoming more capable at a faster pace than people and therefore will effectively outsmart us in a short amount of time,” he says. Professional drivers may be among the first to be displaced by robots but few professions are safe, he says: “There are already robotic journalists. Sure, they aren’t very good but they’re getting better faster than human journalists are.”

Mr Thrun now believes that education is the best way to tackle the big upheavals that are likely to spring from the widespread adoption of artificial intelligence and robotics. But not education as you might know it. “We are still living with an educational system that was developed in the 1800s and 1900s,” he says. “Needs have shifted in the modern age and what’s also shifted is our ability to use digital media. We can now deliver a top-notch education at home in a way that was never possible before.”

The nano alternative

In 2012 Mr Thrun left full-time work at Google to found Udacity, a startup dedicated to reinventing education for the 21st century. Udacity’s “nanodegrees” combine on-demand video lessons, short online quizzes and longer projects, and are designed in collaboration with high-tech companies desperate for skilled workers. The idea is that anyone with a few minutes to spare can log on and work through a programming course at their own pace.

Mr Thrun insists that nanodegrees are distinct from massive open online courses (MOOCs), the digital lecture series which are now offered by many higher educational institutions. Udacity analyses individual students’ learning data (using AI) in an attempt to increase their retention and completion rates. “We effectively reverse-engineer the human learning brain to find out what it means for a person to engage,” says Mr Thrun. “It’s my dream to make learning as addictive as a video game.”

Online lessons and automated tests are free, although students pay for feedback from real humans and to obtain a certificate on completion. Because the teachers are usually recent nanodegree graduates rather than traditional professors, Udacity can keep prices to just a couple of hundred dollars a month, which is about a tenth of the price of a university. Mr Thrun also claims that over 60% of Udacity students finish their courses, compared with around 10% for MOOCs.

Depending on their complexity, nanodegrees are designed to take just 4-12 months to complete. Shorter courses like these are appropriate for today’s high-paced workplace, says Mr Thrun. “The dream of lifetime employment has gone. In my field, whatever you’ve learned becomes obsolete within five years. If you only spend six months on your first degree, as opposed to the average six years for a bachelor’s degree today, you can afford to get more education when you need it again later on.”

Udacity has some 4m registered users worldwide, and about 60,000 students working on nanodegrees at any one time—more in all than the largest university in America. However, at the moment it only offers six nanodegrees, all of which relate to computer programming. Mr Thrun says this reflects a big demand for computer-science courses. He refers to a report by McKinsey Global Institute, a research arm of the consultancy firm, which estimates that by 2020 the world will have 95m more low-skilled workers than employers require, and 85m too few educated workers to fill jobs.

“We have a situation where the gap between well-skilled people and unskilled people is widening,” he says. “Udacity is my response to the development of AI. The mission I have to educate everybody is really an attempt to delay what AI will eventually do to us, because I honestly believe people should have a chance.”

This article appeared in the Technology Quarterly section of the print edition under the headline "Teaching tomorrow"

Washington, we have a problem...

From the September 5th 2015 edition

Discover stories from this section and more in the list of contents

Explore the edition