Cars That Think iconCars That Think

Anthony Levandowski, a former Google engineer and co-founder of the self-driving truck company Otto, which was bought by Uber, in San Francisco, May 16, 2016, in front of an Otto autonomous truck.

By Firing Anthony Levandowski, Uber Got Otto on the Cheap

When Uber bought self-driving start-up Otto back in August 2016, the buzz was not around lidars, patents or trade secrets but the sheer size of the deal. It was widely reported that the acquisition was worth $680 million to Otto’s founders, sweetened by 20 percent of any future self-driving trucking profits. That potential billion dollar pay-out would go to a company barely six months old, and with only a couple of prototype trucks on the road.

The companies did not announce the terms of the deal, but neither did they deny widespread reports that the deal was for 1 percent of Uber’s stock, worth around $680 million at the time.

However, documents filed as part of the ongoing lawsuit between Waymo and Uber now suggest that the ultimate pay-out to Otto could be as low as $220 million–less than one third the original figure.

Otto’s value is important because Waymo has argued in court filings that Uber paying the “remarkable sum” of $680 million for “a company with few assets and no marketable product” is evidence it was really buying misappropriated trade secrets. Waymo sued Uber in February, alleging that engineer Anthony Levandowski took 14,000 confidential technical files from Google’s self-driving car program when he left to start Otto.

Replying in August, Uber wrote: “Waymo’s claims are specious because Uber purchased talent, not technology, when it purchased [Otto] for incentive stock (not $680 million in cash).” But it was only on Monday this week that the actual terms of the purchase were revealed, in a partially unredacted deposition of Uber’s vice president of corporate development, Cameron Poetzscher.

Poetzscher said, “In late 2015, [the Otto deal] was a customer/vendor relationship. And we shifted strategy to say we’d rather have these people as employees... Ultimately, we hired them. We paid $100,000 in cash, and we gave a commitment to issue... approximately 12 million shares of Uber, contingent on meeting milestones.”

With Uber having issued about 1.5 billion shares, the Otto purchase represented 0.8 percent of the company’s stock. That works out to about $566 million at Uber’s 2016 valuation. The cash payment of $100,000 was presumably for each of Otto’s 90-plus employees, accounting for another $10 million at most. The total of $567 million is about $100 million less than the $680 million reported at the time. (Spectrum approached Uber for confirmation of these numbers, but did not receive an immediate reply).

However, a lot has changed since last August. In May, Anthony Levandowski was fired from his position overseeing Uber’s self-driving technologies for not cooperating with the company’s investigation into Waymo’s allegations. At a court hearing, Uber revealed that Levandowski had been granted 5.31 million Uber shares during the Otto acquisition – but also that none of them had vested. When Levandowski left the company, he presumably took with him only the $100,000 signing bonus that all Otto employees received.

Levandowski’s 5.31 million shares accounted for nearly 45 percent of the Otto purchase - and the value of the remainder seem to have plummeted. Although Uber is privately held, its shares can be roughly valued by the amount that institutional investors pay for a stake in the company. The most recent proposed investment, by Japanese technology firm SoftBank, would value Uber’s shares at $33 each – a drop of about a third since its last funding round. That means the remaining 6.69 million shares could be worth just $221 million.

The final total is likely to be smaller still, as key members of Otto may have sacrificed their shares on leaving Uber. Otto co-founder Claire Delaunay is now a vice president of engineering at Nvidia, while other engineers have departed for rival autonomous start-ups Zoox, Starsky, Argo and Scoot.

But what about the promise of Otto employees getting 20 percent of the profits from Uber’s future self-driving trucks? That agreement remains unclear but court filings suggest that it might also have been overstated.

While Uber was negotiating the acquisition in February 2016, Otto split itself into two companies. There was one called Ottomotto that Uber purchased, and one called Otto Trucking that it did not. Otto Trucking, which was originally named as a co-defendant in Waymo’s lawsuit, has provided trucking hardware or services to Uber, and it is this company that appears to have the profit sharing arrangement with Uber.

A heavily redacted draft contract for the acquisition references the trucking company’s profit sharing deal, and Nina Qi, a corporate development manager at Uber, talked about it during a deposition in June. “The Otto Trucking profit sharing pool is an aggregate number,” she said, “And Anthony [Levandowski]’s departure lowered the maximum amount that the team could earn in a collective.”

So firing Levandowski has saved Uber from yet another pay-out, even though any profits from its nascent self-driving trucking business could still be years away.

While Uber’s purchase of the startup seems to have earned Otto far less than was previously reported, it is unlikely that new CEO Dara Khosrowshahi considers it a bargain. The Waymo lawsuit stalled Uber’s development of self-driving technologies, drove away engineers, and continues to reveal a history of shady business practices. Even the cheapest buys can come at a high price.

A slow-motion compiliation of black-and-white images depicting a tennis player. The compilation was produced by TetraVue's flash lidar.

TetraVue Says Its Lidar Will Dominate the Robocar Business

Standard automotive lidars scan objects with moving laser beams, but flash lidar illuminates the entire scene in an instant in order to take a snapshot. This method combines a camera’s high-resolution imaging with a lidar’s range finding, but so far it has been prohibitively expensive.

Now comes TetraVue, of Vista, Calif., with a system that it says can make the lidar’s 2D video feed as high-res as a regular camera’s and at the same time cheap enough to be the preeminent component in the sensor suite of tomorrow's self-driving cars. It uses a pulsed gallium arsenide diode laser as its flash and gauges the distance from each pixel in an array to the object it is imaging. In other words, it turns the 2D image into a 3D one.

Flash lidar has been around for a while, particularly in military applications, where nobody counts the cost. What TetraVue says it brings to the table is a new way of measuring the distance to objects.

“We put an optical encoder between the lens and the image sensor, and it puts a time stamp on photons as they come in, so we can extract range information,” says Hal Zarem, chief executive of TetraVue.

Read More
Hyperloop tubes are displayed during the first test of the propulsion system at the Hyperloop One Test and Safety site on May 11, 2016 in North Las Vegas, Nevada.

Now There's Even a Hyperloop Think Tank

The Hyperloop became a thing some four years ago when Elon Musk first talked about whisking passenger-carrying pods through evacuated tubes at near-sonic speeds. It has since followed the natural progression from notion to concept to idea…to think tank.

The think tank is called the Hyperloop Advanced Research Partnership (HARP), and it means to look into not only the technological but also the economic, legal and social sides of Elon Musk’s proposal. And its preliminary conclusion is that the Hyperloop won’t replace existing systems of transport so much as augment them.

“We have to integrate with rail, trucking and other modes, leaving us with a hybrid mixture,” Dane Egli, president of HARP, said yesterday at a press conference in New York City. “A Hyperloop—over a long distance, at high speed—may, when it gets to the city, link with a bullet train.” And, just as was the case in the early days of rail, some versions of the Hyperloop may be incompatible or barely compatible with others.

Egli and the other members of the HARP board spoke of the Hyperloop’s advantages in roughly the same terms as Elon Musk: Its pods fly frictionlessly on a magnetic cushion, through air under the pressure faced by a jet at 30,000 feet, in a tube that’s all but immune to the vagaries of the weather. It will use computer-controlled movements to stay clear of other pods, and it will find and repair any defects in the depressurization system with static sensors, perhaps aided by robots resembling the “pigs” now used to patrol the insides of natural-gas pipelines. 

There are potential drawbacks as well, as the panel acknowledged. Like a stomach-churning acceleration which passengers will have to endure while strapped into the seats of a windowless pod. But the panel dismissed these as the sort of problems that engineers are paid to solve.

It’s all part of HARP’s dual mission: to promote the technology and at the same time keep tabs on its current status. That’s hard to do this early in the game, when most Hyperloop startups are still guarding their operational information as trade secrets.  “There’s been no comprehensive study yet that has pulled together information on all 10 companies to know what is the state of the technology,” Egli said. (Here’s a rundown on some of the companies.)

However, based purely on information that’s been made public, it would seem that Hyperloop One, formerly known as Hyperloop Technologies, is “the furthest along in technology,” said Chris Zahas, the treasurer of HARP. In any case, Hyperloop One can claim to have the coolest video clips:

Particularly interesting was the discussion of a topic any think tank must entertain: where the money will come from to build an entirely new kind of infrastructure.

“You have to distinguish between financing and funding,” said Prof. R. Richard Geddes, director of Cornell University’s Program in Infrastructure Policy.

Financing, he explained, merely involves the forums in which you borrow money—pension funds, for instance, or maybe sovereign wealth funds, like those of Norway or Saudi Arabia. (Some Hyperloop projects are indeed being entertained in oil-rich Middle Eastern states.)

Funding those loans, by contrast, comes from things like user fees or, perhaps, government subsidies—and there’s no reason why Hyperloop systems can’t use a bit of both, the way so many roads and bridges did.

Getting right of way—perhaps the thorniest problem—should be manageable, Geddes said, because there’s enough space alongside existing highways and railroads. A Hyperloop in the open country would sit on pylons, a configuration that makes for a narrow footprint. And within urban areas the tube could snake through tunnels, some of them cut, perhaps, by Musk’s latest startup, the Boring Company.

It all makes sense, but does it ring true? This think tank is in the tank for the Hyperloop; today’s panel discussed mostly the bright side of Musk’s notion-turned-concept.  Nobody has ever sent people through an evacuated tube, even as a stunt. But even that difficulty can be made to seem manageable.

“If you have a superfast train you need to enclose it,” Geddes said, to keep livestock and misguided people off the track. “For the Hyperloop you just have to make it that much more sealed. It’s not that different.”

A Google Maps Street View car going through the DUMBO neighborhood in Brooklyn, NY.

Deep Learning and Google Street View Can Predict Neighborhood Politics from Parked Cars

It's likely that your car says something about you. The make and model, whether it's foreign or domestic, and how expensive it is can provide information about who owns it. This doesn't work for everyone, of course, but over a large enough population, the statistics can be fairly reliable indicators. The United States spends a quarter billion dollars collecting socioeconomic information by hand through community surveys every year, but if there were a big enough database of what types of cars can be found in which neighborhoods, that data collection could be done more affordably, more frequently, and cover much larger areas. And there is a big database of neighborhood street pictures, in the form of Google Street View imagery. 

Researchers from Stanford University have applied deep learning-based computer vision techniques to 50 million images across 200 regions to identify 22 million cars, which is roughly 8 percent of all automobiles in the United States. Based on the types of cars and their locations, the researchers estimated the income, race, education, and voting patterns of the people living in those areas. The results they derived from pictures are impressively accurate.

Read More
GM and Cruise Automation's Bolt EV self-driving car in San Francisco on November 28, 2017.

GM Demos Robocars in San Francisco

Cruise Automation, the self-driving subsidiary of General Motors, has taken  observers on rides in a more challenging environment than rival Waymo chose for a similar demonstration a few weeks ago.

On Tuesday, Cruise sent a few select journalists through the busy streets of San Francisco. Today, it sent investment analysts as well. Waymo, for its part, conducted its first public rides at a test facility and soon afterward, in the sedate suburban streets of Chandler, Ariz.

Cruise deployed versions of GM’s Chevrolet Bolt, an all-electric car that can drive an impressively long way on a single charge. Waymo’s Arizona pilot relies instead on the Chrysler Pacifica.

Read More
Velodyne's 128 laser beam LiDAR sensor.

Velodyne Unveils Monster Lidar With 128 Laser Beams

Velodyne re-asserted its dominance of the lidar market today by announcing a product with 128 laser beams, twice as many as its previous top-of-the-line model.

“The VLS-128 is the best LiDAR sensor on the planet, delivering the most advanced real-time 3D vision for safe driving,” Mike Jellen, the president of Velodyne LiDAR, said in a statement.

The announcement, which had been widely anticipated, leaves out the one detail that everyone most wants to know: the price. The company’s previous top-of-the-line product originally sold for more than US $70,000.

Read More
An illustration shows a hand holding a stack of books sticking out of a computer.

DARPA Seeking AI That Learns All the Time

Earlier this month a self-driving shuttle in Las Vegas patiently waited as a delivery truck backed up, then backed up some more, then backed right into it. Inconveniently for the roboshuttle’s developer Navya, this happened within hours of the shuttle’s inauguration ceremony. The real problem is that the shuttle can’t learn from the incident the way a human would: immediately and without forgetting how to do everything else in the process.

The U.S. Defense Advanced Research Projects Agency (DARPA) is looking to change the way AI works through a program it calls L2M, or Lifelong Learning Machines. The agency is looking for systems that learn continuously, adapt to new tasks, and know what to learn and when. “We want the rigor of automation with the flexibility of the human,” says the program’s director Hava T. Siegelmann. The US $65-million program has already chosen 16 groups for 4-year projects, but according to Siegelmann there is still opportunity to propose 12- or 18-month projects.

Read More
uber version of volvo xc90 self-driving suv

Uber To Buy 24,000 Self-Driving Volvos

Volvo today announced that it would supply Uber with 24,000 self-driving cars over a three-year period, beginning in 2019. It’s the biggest robocar deal yet.

"Our objective is to be able to operate them without anyone behind the wheel in select cities and environments; the more common definition of that is Level 4 [autonomy]," said Uber's head of automotive alliances, Jeff Miller, in an interview with Automotive News Europe

Miller said Uber chose Volvo in part for its new SPA architecture, which includes wiring that can hook up with today’s advanced driver assistance systems (ADAS) and any further self-driving components that Uber may specify. Uber may, for instance, need additional features in a vehicle meant to be used purely in a ride-hailing service. The electrical system connects all aspects of the car’s functioning—movement, safety, infotainment, navigation—seamlessly.

Such seamlessness does pose certain risks. The company, based in Gothenburg, Sweden, but owned by China’s Geely, has to ensure that the car’s electronic backbone is rock-solid and resistant to hacking. Still, Volvo—perhaps more than any other car maker—has emphasized the importance of multiple redundant safety systems.

The model in question is the Volvo XC90 crossover, which can seat up to seven people. Presumably that means seven passengers and no driver, although who knows, maybe the service will start with a safety driver. That’s what Waymo is doing right now in its ride-hailing pilot program in Chandler, Ariz. However, this month Waymo took the safety driver from behind the steering wheel and put him into the back seat.

Volvo plans to begin its own self-driving program with a fleet of 100 XC90s in Gothenburg, Sweden, beginning next month.

Neither Volvo nor Uber put a price on the project or a precise date for its rollout. The technology would first have to be ready, of course, but so would new rules of the road, at least in the districts where the ride-hailing service is to operate.  

A big rig and a sports car point Tesla in the direction Elon Musk has envisioned

Tesla Launches an Electric Semi Truck—and a New Sports Car

Elon Musk unveiled Tesla’s fifth—and sixth—vehicles at a characteristically glitzy event in Los Angeles last night. An audience of potential customers and current Tesla owners witnessed smooth acceleration, excellent cornering and impressive performance—and those were just the exploits of the electric Semi truck they had turned up to see.

The crowd was then treated to the surprise unveiling of what is destined to be not only the world’s longest-range electric car, but also the fastest production car of any flavor: the new Tesla Roadster.

The unveiling of the two vehicles amounted to a triumphant evening for Elon Musk, who has been struggling through what he admits has been “production hell” with ongoing delays to his company’s mass market Model 3 car.

Although the new Roadster is destined to grab headlines, it was the launch of the Semi heavy-duty truck that could have the most impact on the world’s roads. The truck, which is powered by a Model 3 motor on each of its four drive wheels and boasts a drag coefficient lower than a Bugatti Chiron, can drive 800 kilometers (500 miles) on a single charge of its lithium-ion battery pack.

Tesla would not confirm the pack’s capacity or chemistry, but did say that it is very similar to another product sold by Tesla (probably either the Powerwall domestic or Powerpack utility-scale battery). It says the Semi weighs roughly the same as existing diesel trucks fully loaded with fuel.

The pack will be able to take on 650-km-worth of charge in 30 minutes at a new generation of charging stations, called Megachargers, that Musk will soon roll out to Semi drivers operating “anywhere in the world.” The Megachargers will deliver solar power generated from Tesla solar panels, selling it at a flat rate of 7 cents per kilowatt-hour in the United States. That is just over half the national average rate of 13.1 cents per kWh.

At that price, says Musk, the Semi will cost just 78 cents per kilometer ($1.26 per mile) to operate. The American Transportation Research Institute (ATRI) says that operating a diesel truck in 2016 cost an average of 99 cents per km ($1.59/mi). That differential means that the Semi will save truckers hundreds of thousands of dollars, despite it having a higher purchase price than today’s diesel trucks. Musk did not specify a retail cost for the Semi, but Jerome Guillen, the engineer in charge of the Semi project, did say Tesla already has “a lot” of preorders.

Trucking may seem like a departure for the Californian company more commonly associated with luxury cars like the Roadster announced later. But it actually fits neatly into Musk’s long-term strategy of covering all “major forms” of terrestrial transport, with the aim of accelerating the advent of a sustainable energy economy.

According to the International Energy Authority, trucks account for around 20 percent of global oil demand, 35 percent of transport-related carbon dioxide (CO2) emissions, and around 7 percent of the planet’s energy-related CO2 emissions. If trucks continue to burn oil, the IEA predicts their oil consumption will rise by nearly one-third again by 2050. Instead, says Musk, “Our trucks run on sunlight."

Tesla says the Semi comes with the same Enhanced Autopilot system as its latest cars, using identical front- and side-facing cameras, computers, ultrasonic sensors and radars—although probably more of them. As expected, it does not have a lidar system. Musk did not talk about the potential of automating drivers out of a job, instead focusing on the safety and efficiency benefits of partial autonomy.

A platooning convoy of three Tesla Semi trucks, he said, would reduce their cost per kilometer to just 53 cents (85 cents/mi). This would be cheaper than sending the same freight by train. “This is economic suicide for rail,” he said.

Tesla is also keen to woo drivers by putting them in a stylish cab that is home to a single, centrally-located seat with 15-inch touchscreens on either side. These come straight from the Model 3, displaying a similar selection of driving information and maps, and access to settings and audio entertainment. In fact, many of the Semi’s components are carryovers from other Teslas, including exterior door handles from the Model 3. But unlike the Model 3, it does not have a driver-facing internal camera.

While the Semi promises cheaper, more efficient freight transportation, the new Roadster is unabashedly presented as a shiny toy for rich speed freaks. Its ability to punch from 0 to 100 km in less than two seconds and reach a top speed of over 400 km per hour, makes it the spiritual successor to the Lotus-bodied Roadster that launched Tesla. With a range of 1000 kilometers (620 miles), it can zip from Los Angeles to San Francisco and back again on a single charge.

The Roadster’s $250,000 price tag is about the same, probably, as the Semi truck. But only one quarter-million-dollar investment has the potential to slash carbon emissions, eliminate air pollution and even reduce accidents.

The Terrafugia flying car is seen in the air, with roads below.

Flying-Car Company Terrafugia is Bought by China's Geely

“We wanted flying cars, instead we got 140 characters," wrote tech billionaire Peter Thiel, in 2011. Well, Twitter is now going to allow 280 characters in a tweet, and soon—we are told—we will finally get a flying car.

Terrafugia, which in 2006 became one of the very first of a new crop of flying-car companies, says it will have one in 2019. And though that is only the latest in a long string of deadlines the company has set, perhaps—Perhaps!—this will be the first one that it honors. Reason: The company has just been acquired for an undisclosed sum by Geely, the Chinese automotive company that owns Volvo. The R&D funds should flow profusely.

A lot of companies from around the world are chasing the same dream—among them Toyota, Google’s Larry Page, Germany’s eVolo, and China’s EHang. Vahana, a subsidiary of Airbus, was supposed to test-fly a prototype this year, as I wrote back in January, but here we are in November with no air taxi. I am not shocked, that this is the case; in this business, deadlines are aspirational.

What’s more, the Vahana craft was supposed to be autonomous. That makes sense, even if it’s devilishly hard to manage. There’s just no other way to make money off a battery-powered plane that can ferry a mere one or two passengers at a time.

Terrafugia has dreams of autonomy as well, in the shape of its TF-X model, but the company has put its deadline a good seven or eight years further out. Not including a fudge factor.

Okay, so it may seem we’re beating up on a futuristic company that dares to dream big. But IEEE Spectrum and Terrafugia go back a long way, like an old married couple. We’ve been grousing about why we consider Terrafugia’s ambitions unreasonable ever since we first labeled the company a “loser,” back in 2007.

The reasons for our skepticism haven’t changed. Making a car into a plane is really just a way of making a plane that can kinda, sorta move down a road without knocking off the heads of parking meters. Such “roadable” planes have absolutely no appeal as a car. Indeed, the same can be said of those crazy cars that were supposed to float like boats or even submerge like submarines. No, seriously! Take a look at the Lotus Esprit that sent James Bond diving deep, in “The Spy Who Loved Me” (1977).

One thing has changed, though. Today there’s a lot of serious money behind the flying-car dreams, and the question is why.

It could be sheer personal enthusiasm. For instance, Billionaire Larry Page is putting his own money into his secretive flying-car project. Maybe there just wasn’t a good enough business case for Google to back it with corporate funds.

Then there’s the tech-demo argument. Doing hard things on a small scale may be a good way to master robotics, sensor fusion, and other skills that will come in handy in self-driving cars and super-automated jetliners. This argument may lie behind Geely’s acquisition policy, which has also swept up companies with expertise in other aspects of self-driving cars and suchlike. Here’s a checklist of the tech moves the Chinese company has made lately, from Automotive News.

Finally there’s the cool factor. Elon Musk—who, interestingly, is not trying to give cars wings—has always used outlandish projects to fire up the imagination. Witness the Hyperloop, the Boring Company, SpaceX—and Tesla itself. Such quests for miracles attract top engineering talent.

But remember: Musk originally got rich—alongside Peter Thiel—by developing a new way to handle money, culminating in PayPal. If you’re looking for a commercially compelling tech project, look no further.


Cars That Think

IEEE Spectrum’s blog about the sensors, software, and systems that are making cars smarter, more entertaining, and ultimately, autonomous.
Contact us:

Philip E. Ross
New York City
Willie D. Jones
New York City
Senior Writer
Evan Ackerman
Washington, D.C.
Mark Harris

Newsletter Sign Up

Sign up for the Cars That Think newsletter and get biweekly updates, all delivered directly to your inbox.

Load More