Udacity spins out self-driving taxi startup Voyage

Image jpg

UPDATE: Sebastian Thrun answers student questions for 24 minutes in this video Self-Driving Car Nanodegree: Q&A with Sebastian Thrun. This video is probably the most informative insider perspective on the fast-moving autonomous vehicle space.

One example of how fast the field of AI is moving: Udacity’s “school for robo-cars has been so successful that it’s now spinning out of Udacity into its own company, Voyage.” Here’s a snippet from Business Insider:

(…snip…) The new spin-out will be lead by Oliver Cameron, a Udacity VP that was spearheading a lot of its self-driving car curriculum. The company broke the news to its employees Wednesday morning.

Udacity will have a stake in the newly-formed company as part of the deal, said the Udacity’s CMO Shernaz Daver. Voyage also recently closed a seed round of funding that included Khosla Ventures, Initialized Capital, and Charles River Ventures.

Voyage has been hot in Silicon Valley investor circles because of one big name linked to Udacity: Sebastian Thrun. Thrun, who founded the education startup, is also nicknamed the “Godfather of self-driving cars” for the work he did at Google and helped launch the self-driving car nanodegree program at Udacity.

Thrun, though, says he’ll have no connection with Voyage even though it’s spinning out of his company. “Because of personal conflicts, I have excused myself from any involvement in Voyage. I wish Oliver and his team all the best,” Thrun said in a statement to Business Insider.

The autonomous taxi startup wants to bring about the end goal where autonomous cars can carry people anywhere for a very low cost, Cameron said. It already has permission to deploy its self-driving cars to ferry passengers in a few places over the next few months, but Cameron declined to specify where.

“We want to deploy these not within five years, but very soon. We think in terms of weeks, not in terms of years or months,” he told Business Insider in an interview.

Pure guess: one reason Oliver Cameron decided to take this risk is because Udacity is open-sourcing it’s own self driving car project. All the code is there for all of us to use and improve. Including the new startup Voyage. And Oliver Cameron has a pretty good idea how successful the Udacity project is going to be.

Update: this week BMW has announced they plan to ship self driving cars in four years, in 2021. That’s similar to plans already announced by GM, Ford, Chrysler, Mercedes, Volvo and Chinese ride-sharing giant Didi Chuxing.

2017 Udacity Intersect: the future is closer than you think

We meat-bodies generally over-estimate short-term progress (one to three years) and underestimate medium-term technology progress (ten to twenty years). My particular interest is AI. That is partly because five decades ago Artificial Intelligence was my academic field at Carnegie Mellon. The logic-based AI that I was investigating with Herb Simon and Allen Newell is now known as GOFAI (good old-fashioned AI, that’s the AI that didn’t work very well). What motivates my current interest is that Machine Learning (ML), is starting to be really useful, and the rate of progress in narrow AI applications of ML is accelerating. You can see for yourself the rate of progress in ML in the explosion of speech recognition gadgets like Amazon Alexa, the voice service that powers the home Echo device. Now any of us can access voice and image recognition. In perhaps five years we will begin benefiting from Self Driving Cars (SDC) if we live in the right places.

I think that open-source, low-cost initiatives like Google’s release of TensorFlow mark a major inflection point in the rate of ML progress. A teenager can now prototype her ML-based idea to see if it really works. She doesn’t need to go to Sand Hill Road to raise venture capital. In fact, the VC community isn’t going to pay any attention to you unless you have already built your project to a level where it can be tested.

Another powerful indicator of the inflection point thesis is Udacity’s Nanodegree offerings. Outstanding example: the Self-Driving Car Engineer Nanodegree. For $800/term you can graduate with a qualification that dozens of leading companies are eager to hire. Here’s some of the hiring-partners of this SDC Engineer Nanodegree [they helped build the course]:


How can you learn more? Well, I suggest having a look at videos from the March 8th 2017 Udacity Intersect conference. This was a remarkable event, opening the window so all of us can get visibility on what is happening. The Computer History Museum was vibrating with the energy of companies recruiting Udacity students and Nanodegree graduates. Pretty much every panel and keynote of the Agenda was packed with tech industry insiders exchanging views about their projects, priorities and especially the people they want to hire.

I’m highlighting Udacity Intersect 2017 because the conference offers a concise and fun way to get a look into the future and behind the curtain. What are the leading tech companies thinking and doing? Where are we likely to be in 10 to 20 years?

You can find links to videos of every segment of the conference at the main Intersect page. If you’re not sure this is for you, please check out the final session Fireside Chat: Astro Teller and Sebastian Thrun. These 33 minutes gives you access to two of the leading innovators who have convinced me that “the future is closer than you think”.


Update: Udacity spins out self-driving taxi startup Voyage. Also, this week BMW has announced they plan to ship self driving cars in four years, in 2021. That’s similar to plans already announced by GM, Ford, Chrysler, Mercedes, Volvo and Didi.

Morgan Stanley predicts driverless car ‘utopia’ by 2026

Morgan Stanley Research

Adam Jonas isn’t Nostradamus, but the Morgan Stanley analyst is predicting the road to complete vehicle autonomy will begin in 2026. What’s more, he says the technology will eventually reach 100 percent market penetration two decades later.

Wall Street forecasts aren’t generally superior to the dart-throwing monkeys, but they do create great charts. At end-2018 we can reality-test the Adam Jonas report: there should be at least a couple of full-autonomous offerings on the market.

The majority of media reports I see on robocars are focused on the idea that Danny-the-driver sells his Prius, buys his first robocar — but continues to drive to work every day. The “big change” is that now Danny can TXT while not-driving instead his usual TXTing-while-driving, otherwise known as DWD, Driving While Distracted. That will save a lot of lives, including those of cyclists like myself.

But that isn’t the significant revolution. The real revolution will be most obvious in high density cities like New York or San Francisco. High robocar penetration will happen there first because of the appeal of “whistle-cars”. We are already seeing this in the explosive growth of Uber, Lyft and similar services. Urbanites are proving they prefer to summon a just-in-time ride on their iPhone. Robocars will make this service even more convenient and a LOT cheaper. That’s pretty much the end of the self-owned urban car market.

Even earlier than the whistle-car we will see delivery-bot vehicles. This is where Amazon is going with their drone development program. The drones will be the “last block” of the delivery web. The delivery-bots will handle the larger, heavier delivery loads and supply the drones and the human powered deliveries.

American cities typically have 40 to 50% of their useable area eaten by automobiles, comprised of streets and parking. Most of that space can be released to productive use once robocars and delivery-bots reach full penetration. Perhaps cities will even allow building again – so the acute shortage of affordable housing can be eliminated. Or maybe not – the same old status-quo people will probably still control the city governments. And they already have their multi-million dollar positions – no “housing crisis” for them.

For a deep dive into the implications of robocars, I recommend Brad Templeton, now a consultant to Google’s robocar program. See Brad’s main robocar page: Where Robot Cars (Robocars) Can Really Take Us. There’s lots more here on Seekerblog

Paul Saffo on Google’s self-driving cars

Paul speculates on the implications of the Google/Stanford research. There’s a short video of one of the Google robo-Prius cars on the highway.

(…) Why buy a car when you can subscribe to one? This is the alluring premise behind services like Zipcar and City CarShare, but thanks to an assortment of small annoyances, car sharing is a tiny industry that remains more promise than reality. Advanced telematics could be what makes car-sharing practical by adding features calculated to protect occasional drivers and the cars they share. Collision-avoidance systems would spare drivers the grief of an accident and allow for lower insurance rates. Cloud-based mapping and traffic alerts would dramatically increase the convenience of shared cars and provide better monitoring and diagnostics for the system operators. Thrun even thinks we can create a service where instead of picking up your car in a nearby parking space, the driverless car will come to you like a well-trained spaniel.

Stanford self-driving cars: roundtable interview

I just listened to the 40 minute audio podcast of a fascinating roundtable discussion. The key people are Sven Beiker, executive director of the Center for Automotive Research at Stanford, and futurist Paul Saffo, managing director of foresight at Discern Analytics. There is also video of the Reporters’ Roundtable discussion.

Paul Saffo contributed several new insights on how robotic cars are likely to evolve. Already in Japan and German the twenty-somethings do not want to own a car. E.g., consider the emergence of a vastly larger market for shared cars when they become all electric robots with predicted lower insurance rates. Google might see this as a great way to create almost real-time Street View coverage by adding the camera to all these cars.

Paul also predicted that the US military will be an early adopter for convoys: human driven lead vehicle followed closely at high speed by the robot vehicles. Consider the implications for countering attacks on the convoys.

Show notes and talking points

Overview: Where are we on the road to the autonomous car?

Talk about current autonomous solutions and driver aids.

What’s in robocars for Google?

Problems yet to solve, like legal and insurance issues. Who gets the ticket? Who’s responsible for a crash?

What happens when the first robot car crashes and kills its driver or a bystander?

If the human is ultimately responsible, how are you going to keep them alert when the car is doing all the work?

Will we need new roads? New driver education?

Talk about how contests, challenges, and racing move innovation forward. See the Grand Challenge (PDF).

Talk about Pikes Peak car at Stanford. What’s the difference between it and the DARPA Challenge car, Stanley?

How will robocars come to real-world car buyers? When can we buy them?

More on Google’s self-driving car program

Stanford’s Sebastian Thrun is my hero for steering the Stanford DARPA Urban Challenge project to success. Thrun is one of the hired-guns on Google’s high intensity project (think a bit on the technologies that might enable robo-cars: the synergy of AI + mobile + localization + GIS database). I am very curious how much of the Google robo-car algorithms depend on real-time data coming down from their cloud. Nobody can match Google’s cloud for auto-driving:

(…) So we have developed technology for cars that can drive themselves. Our automated cars, manned by trained operators, just drove from our Mountain View campus to our Santa Monica office and on to Hollywood Boulevard. They’ve driven down Lombard Street, crossed the Golden Gate bridge, navigated the Pacific Coast Highway, and even made it all the way around Lake Tahoe. All in all, our self-driving cars have logged over 140,000 miles. We think this is a first in robotics research.

Our automated cars use video cameras, radar sensors and a laser range finder to “see” other traffic, as well as detailed maps (which we collect using manually driven vehicles) to navigate the road ahead. This is all made possible by Google’s data centers, which can process the enormous amounts of information gathered by our cars when mapping their terrain.


Read more »

Google self-driving cars

Charles just sent us the link to this NYT Science article on Google’s self-driving car program. We’ve been following the Stanford-Volkswagen program — I did not know that Google was participating. Is Sebastian Thrun now a Google employee? Sebastian lead Stanford’s “Stanley” DARPA Grand Challenge project.

(…) During a half-hour drive beginning on Google’s campus 35 miles south of San Francisco last Wednesday, a Prius equipped with a variety of sensors and following a route programmed into the GPS navigation system nimbly accelerated in the entrance lane and merged into fast-moving traffic on Highway 101, the freeway through Silicon Valley.

It drove at the speed limit, which it knew because the limit for every road is included in its database, and left the freeway several exits later. The device atop the car produced a detailed map of the environment.

The car then drove in city traffic through Mountain View, stopping for lights and stop signs, as well as making announcements like “approaching a crosswalk” (to warn the human at the wheel) or “turn ahead” in a pleasant female voice. This same pleasant voice would, engineers said, alert the driver if a master control system detected anything amiss with the various sensors.

Volkwagen, Stanford: the future of autonomous vehicles

…are working together to make Driverless Cars a reality in a cooperation that includes the new VAIL facility on the Stanford campus (Volkswagen Automotive Innovation Laboratory), the opening is pictured above, and the VERL (Volkswagen Electronics Research Laboratory in Palo Alto).

Volkwagen is very serious about driverless vehicle technology. The TT-S in the photo above is being prepared to enter the Pike’s Peak International Hill Climb in August (yes, I said driverless!). Here is a fun VW video made to publicize the venture with Stanford.

Volkswagen and Stanford University go way back, having collaborated on two cool autonomous cars for DARPA, and now they’ve gone in together on a laboratory where researchers and students will develop technology they say will lead to safer, greener cars.

The German automaker, through Volkswagen Group of America, is investing $5.75 million in the Volkswagen Automotive Innovation Laboratory to spur the creation of new automotive tech. Along with VW’s Electronics Research Laboratory in Palo Alto, VAIL gives VW the largest Silicon Valley research presence of any automaker.

“This collaboration can draw on a long-standing relationship between the Volkswagen Group and Stanford, which continues to increase the exchange between industrial and academic talent,” Dr. Franz-Josef Paefgen, chairman and CEO of Bentley Motors, said in a statement. (VW owns Bentley.) “The goals are to accelerate automotive-related research on campus, increase opportunities for collaboration between the VW Group and Stanford and build a global community of academic and industrial partners committed to the future of automotive research.”

VW has been working with Stanford since at least 2005, when the two collaborated on Stanley, an autonomous Touareg that won the DARPA Grand Challenge that year. Stanley is now on display at the Smithsonian. It was followed in 2007 by Junior, a Passat that was runner up in the DARPA Urban Challenge. That’s Junior in the main pic, being driven sans hands.

Volkswagen, through Audi, also has developed an autonomous TT-S, pictured below, that will attempt to conquer Pike’s Peak sometime next year. That’s an audacious goal, given the 12.42-mile course to the 14,110-foot summit features 156 turns and is among the greatest challenges in motorsports.

Burkhard Huhnke, Director of Research for Volkwagen America was the second speaker in the Cato event Driverless Cars: The Next Transportation Revolution. Huhnke described the Stanford venture, inviting us to visit the center, see the self-parking cars, etc.

And for some adrenalin, try this professional video of one of the top Ford teams climbing the peak in an 800 horsepower Ford Fiesta Rallycross:

In order to reach the summit, drivers must negotiate 12.42 miles of winding mountain road, there is imminent danger awaiting just around each one of the course’s 156 turns and while rising the 4,721 feet to the peak, drivers will have to adapt to everything from paved highways surrounded by pines to loose gravel surrounded by nothing more than wide open blue Colorado skies.


The vehicles will be heavily modified rally cars based on the Fiesta hatchback, and will be powered by 2.0 Liter Duratec Ford four cylinder engines, capable of putting out over 800 HP each. Ford will bring Swedish Rally Champion Andreas Eriksson and World Rally legend Marcus Grönholm to try and conquer the mountain.

This is comparatively tame back at VAIL on the Stanford campus: venture capitalist Steve Jurvetson posted this short video of VW Junior 2 self-parking at Stanford.

The DARPA Urban Grand Challenge is upon us…

The next exciting chapter in the DARPA robot-vehicle challenge begins Oct 26th [qualifier] and Nov 3 [the race assuming there are successful qualifiers]. Wikipedia has a good set of resource links and a good summary of the urban challenge:

…For 2007, DARPA introduced a new challenge, which it named the “Urban Challenge”. The Urban Challenge will take place on November 3, 2007 at the site of the now-closed George Air Force Base (currently used as Southern California Logistics Airport), in Victorville, California (Google map).[2] The course will involve a 60-mile (96 km) urban area course, to be completed in less than 6 hours. Rules will include obeying all traffic regulations while negotiating with other traffic and obstacles and merging into traffic. While the 2004 and 2005 events were more physically challenging for the vehicles, the robots operated in isolation and did not encounter other vehicles on the course. The Urban Challenge requires designers to build vehicles able to obey all traffic laws while they detect and avoid other robots on the course. This is a particular challenge for vehicle software, as vehicles must make “intelligent” decisions in real time based on the actions of other vehicles. Other than previous autonomous vehicle efforts that focused on structured situations such highway driving with little interaction between the vehicles, this competition will operate in a more cluttered urban environment and requires the cars to perform sophisticated interactions with each other, such as maintaining precedence at a 4-way stop intersection.

Stanford’s team “won” the 2005 Grand Challenge along with three other successful finishers [Terramax also finished, but 12 hrs vs. 10-hour maximum time]. NOVA did a special on that event, available as DVD or online. The 2007 Stanford effort offers a video of their new “Junior” demonstrating its competence.

You can tell these are the “good guys” by their laptop selection.

Possibly the most remarkable part of the Grand Challenge story is the success of #4 finisher Team Gray, some 37 minutes behind Stanley. This is a privately funded New Orleans insurance company, with Tulane student tech help. And they lost 3-4 weeks from an abbreviated development schedule due to Katrina [thus their vehicle’s name “Kat-5”]. Team Gray has qualified for the Urban Grand Challenge as well.

…In an unheralded fourth place, though — just 37 minutes behind the winner — was an entrant from Gray Insurance, a small, family-owned casualty company in Metairie. Fourth place may not sound like any big whoop until you realize that the Gray Team finished ahead of cars from some of America’s most elite technical universities, as well as from a number of big defense contractors. In fact, theirs was one of only five vehicles that managed to even cross the finish line.

Since Mr. Gray admits to knowing virtually nothing about computers, he turned things over to Keith B. Goeller, who heads Gray’s 10-member IT department. The plan, says Mr. Goeller, was to keep his team small, look for the best off-the-shelf products, and then assemble them without being religious about anything.

…Mr. Gray says the project cost about $650,000, and was supposed to be good PR for the company, though the ongoing hurricane mop-up has kept the insurers from making hay from the accomplishment.

The moral of the story?

“It’s a beautiful thing when people are ignorant that something is impossible,” says Mr. Gray. “In fact, that’s the American way.”

But for what they call “the $2 million dollar bug” Team Gray might have beaten Stanley — their software slowed the vehicle on the flat lake bed segments due to processing overload attempting to scan the very broad possible road area.

While searching for resources on the DARPA prize I came across a YouTube video of one of the two-wheeled 2005 challengers. The video neatly illustrates how one can correct an impending bicycle fall by turning into the direction of fall. Very handy if you are caught with your foot trapped in a foot-binding — at least if your reflexes can implement the technique before it is too late…