Street View Data Not Good Enough For Google's Self-Driving Systems

As reported by AutoWeek: If anyone can bring the self-driving car to market, it’s Google. Not only does the company have the necessary funds for such an undertaking, it has the necessary data -- at least one would think.

In a Wired article on Tuesday, Google said that even though it has mapped much of the world with Google Street View, those maps won’t help its self-driving car. The vehicle needs to relearn the road before it can tackle driving autonomously.

Currently, the self-driving test SUVs have mapping hardware, including a laser scanner that rotates 10 times per minute, with 64 beams measuring the distance to nearby objects. The camera and radar bring in more information, which is compiled into a picture of what’s going on. All of this data is necessary for the autonomous function to work.

The solution, as Google puts it, is that each car will have the systems required for mapping a road. The company foresees an owner who lives in an unfamiliar area driving the car for the first few times until the vehicle learns the street. After that, it could go autonomous.

The U.S. has approximately 4 million miles of public roads. Google would need to tackle much of that if it wants to meet the goal of co-founder Sergey Brin, who said in 2012 that he wants to have the technology commercialized by 2017.

In the Future, Driverless Cars Could Cripple Law Enforcement Budgets

As reported by NetworkWorld: Shortly after the state of Washington voted to legalize recreational marijuana late last year, opponents made a very interesting, if somewhat counter-intuitive  argument against legalized pot – law enforcement would miss out on the huge revenue stream of seized assets, property, and cash from pot dealers in the state.

Justice Department data shows that seizures in marijuana-related cases nationwide totaled $1 billion from 2002 to 2012, out of the $6.5 billion total seized in all drug busts over that period. This money often goes directly into the budgets of the law enforcement agencies that seized it.

One drug task force in Snohomish County, Washington, reduced its budget forecast by 15% after the state voted to legalize marijuana, the Wall Street Journal reported in January. In its most fruitful years, that lone task force had seen more than $1 million in additional funding through seizures from marijuana cases alone, according to the report.

Naturally, this dynamic is something law enforcement either is or should already be preparing for as driverless cars make their way onto the roads. Just as drug cops will lose the income they had seized from pot dealers, state and local governments will need to account for a drastic reduction in fines from traffic violations as autonomous cars stick to the speed limit.


Google’s driverless cars have now combined to drive more than 700,000 miles on public roads without receiving one citation, The Atlantic reported this week. While this raises a lot of questions about who is responsible to pay for a ticket issued to a speeding autonomous car – current California law would have the person in the driver’s seat responsible, while Google has said the company that designed the car should pay the fine – it also hints at a future where local and state governments will have to operate without a substantial source of revenue.

Approximately 41 million people receive speeding tickets in the U.S. every year, paying out more than $6.2 billion per year, according to statistics from the U.S. Highway Patrol published at StatisticBrain.com. That translates to an estimate $300,000 in speeding ticket revenue per U.S. police officer every year.

State and local governments often lean on this source of income when they hit financial trouble. Last September, Atlanta’s Channel 2 Action News reported that the city’s police union chief claimed in an email to city police officers that Atlanta Mayor Kasim Reed had “designated traffic court/ticket revenue for future pay increases” for police officers. There are countless other examples of police officers increasing their ticket-writing efforts in order to help spike revenue.


And a study released in 2009 examined data over a 13-year period in North Carolina, finding a “statistically significant correlation between a drop in local government revenue one year, and more traffic tickets the next year,” Popular Science reported.

"If a county got one percentage point less money, they gave out around a third of a point more tickets," according to the Popular Science report.

Of course, it’s unclear what exactly the driverless car ecosystem will look like. Tesla CEO Elon Musk, who says his company's autonomous car will be ready for on-street adoption by the masses by 2016, doesn’t believe consumers will adopt fully autonomous cars, but rather those with "a form of 'auto-pilot' in most situations that would allow the vehicle to take over control."


"My opinion is it’s a bridge too far to go to fully autonomous cars," Musk told the Financial Times last September. "It’s incredibly hard to get the last few percent."

Musk makes a good point. A car that can drive you home when you’re tired or have had too much to drink seems like a great idea, but consumers may not be crazy about being restricted to speed limit-abiding robots when they’re running late for a meeting or about to miss a flight. This could also explain why Google has reportedly had difficulty finding a manufacturing partner for its fully autonomous car, as the Financial Times pointed out in its article about Musk.




The question, however, is how much of an impact any autonomous driving ability has on the reliable revenue source that traffic violation fines have been in the public sector. If and when drivers decide to switch off their autonomous drivers and double the speed limit on a joy ride, will police be able to pay enough officers to stop them?

SpaceX Has Gone From Suing The Air Force To Possibly Launching Its Satellites

As reported by Business Insider: SpaceX, billionaire Elon Musk's aerospace startup, is reportedly taking its first steps towards securing lucrative U.S. defense contracts. 

Bloomberg News reported today that the Air Force — which SpaceX sued over anti-competitive practices as recently as last month — now plans on spending $60 million to certify SpaceX for carrying its payloads.

The certification process will likely involve close analysis of each launch of SpaceX's Falcon 9 rocket, as the Air Force tries to determine whether it thinks that SpaceX is competent to carry its payloads. The Air Force will be looking at factors like the Falcon 9's launch environment and at whether the rocket, which is built to be reusable, can still deliver payloads if some of its engines fail in-flight.

That SpaceX is now on track for military contracts is a sign of the changing nature of the U.S. space industry. As Kieth Cowing, a former NASA astrobiologist and current blogger at NASA Watch told Business Insider, the U.S. government likely never thought it would have to certify another potential launch company because of larger companies' long-standing dominance of aerospace-related government contracts. "The government did not expect to ever have to do this again, because you've got a sanctioned monopoly with Boeing and Lockheed Martin," says Cowing.

SpaceX reusable rocket tested to 1000 meters.

SpaceX has been trying to change that. Last month, the company sued the Air Force over its allegedly non-competitive process for awarding launch contracts. And earlier this week, SpaceX's Dragon capsule, a reusable cargo module, successfully returned to earth from the International Space Station.

The company's emergence, and the government's sudden openness to aerospace newcomers, comes at an opportune time: Both Boeing and Lockheed rockets use engines built in Russia. SpaceX's delivery systems are built and designed in the U.S. — so its work isn't potentially hamstrung by Russia's aggression in Ukraine and increasing alienation from the U.S. and its allies.

Rogue Google X Internet Balloon Initially Identified as a UFO

As reported by Motherboard: The hardest thing about Google X's Project Loon hasn't been the engineering challenge of beaming high-speed internet down to the far-flung corners of the world from 60,000 feet up in the sky, explained Richard DeVaul, the founder of the moonshot project. It's trying to control all those freaking balloons.

DeVaul, sporting a pair of Google Glass onstage at the Smithsonian's "Future Is Here" event this weekend, gave us a rare behind-the-scenes glimpse at one of the secretive lab's most high-profile snafus.

"I can tell you about one of the unexpected things that happened," he said. "We created mass UFO sightings across the US."

The most famous of these occurred two years ago in Pike County, Kentucky. Alien-minded amateur astronomers spotted and captured footage of an aerial object unlike anything they'd ever seen before floating high up in the sky.

"I don't understand how it could stay up there in one place. There was no sign of propellers or any kind of proportion system. No gas is coming out," wrote Allen Epling, who first spotted the flying object. The airline industry reported they didn't have any planes in that area, and police were equally stumped.

After much speculation, conspiracy theories, and subsequent sightings in Virginia and Tennessee, Google admitted the mysterious floater was one of the balloons they were testing for Project Loon. The UFO community wrote the claim off as a publicity stunt, and remained skeptical.

On Saturday, DeVaul revealed to the public for the first time the once-unidentified flying object. It was the "Falcon 11," a 120-foot long transparent mylar balloon made in-house at the secret Google X lab.

"This is a balloon that went rogue," DeVaul said.

The Loon team launched the massive spherical, internet-carrying balloon from California on what was supposed to be a short-duration test flight. But it "ran out of juice" and they lost track of it.

They then followed its course eastward by checking in on UFO forums online.

"We tracked the balloon by outsourcing to the internet UFO community," DeVaul said. "It drifted all the way across the country, getting to Kentucky, to the Eastern Seaboard, all the way up to Canada."

"No one outside Google X knows this."

I caught up with DeVaul after the event and asked him if the Falcon 11 was still in use, or if that particular model was benched after its extraterrestrial disappearing act.

Turns out it was one of many models. All the balloon prototypes are named after birds, starting with "A," he told me. So that gives you some sense of how many attempts there have been to get the design right—and the "Falcon" is from back in 2012. 

When I asked DeVaul what letter they're on now, he laughed and said unfortunately he wasn't able to talk about that (surprise surprise), but suffice to say they were a lot further along. He then hinted that the lab is now working on totally something new—he wouldn't confirm if it's even a balloon—that, if successful, could stay aloft forever.

Typical Google X vaguery and dramatic flare. We’ll be on the hunt for more details on that mysterious object, and in the meantime, the Project Loon team will ostensibly keep trying to perfect its balloon design. Note to UFO chasers everywhere: It might not be an alien spacecraft, it might just be an internet-carrying “Toucan” balloon gone rogue.

Streaming Video and Premium Content Behind AT&T's Buy of DirecTV

As reported by The Verge: Sunday's news that AT&T had agreed to acquire DirecTV for a whopping $48.5 billion came as no surprise to observers of the pay-TV industry. "If you think back to the ’90s the marketplace was full of small companies. We've seen wave after wave of mergers and now there are fewer and larger companies," says Jeff Kagan, an independent analyst. "Going forward were going to see even fewer and even larger competitors going forward or moving toward a national, competitive marketplace for television, telephone, internet, wireless."  

For AT&T, the deal is mainly about gaining scale in video and acquiring the bargaining power that comes with that to license premium content — particularly with the looming specter of a tie-up between Comcast and Time Warner Cable. AT&T will combine its 5.7 million U-verse TV customers with DirecTV's roughly 20.3 million US subscribers. "All of a sudden you're talking about the number-two pay-TV provider in the country," says Dan Rayburn, an analyst with Frost & Sullivan. "That means you can negotiate for better programming, and at a better price."

Having access to premium content is key to AT&T's ambition to become a major player in the world of streaming video. "DirecTV is way ahead of AT&T in terms of licensing deals. Something like NFL Sunday Ticket is a game-changer for AT&T if they can offer it as part of a package to their wireless customers," says Rayburn. AT&T CEO Randall Stephenson said on an investor call this morning the deal will allow AT&T to offer premium video on all screens, from TVs to smartphones to cars and airplanes.

Just how important is big-ticket content like the NFL? The deal terms actually stipulate that AT&T can walk away from the merger if DirecTV doesn't win the contract with the NFL to renew that exclusive.

There are synergies on the DirecTV side as well. Currently DirecTV can't offer its customers competitive high-speed internet as part of its satellite package. "AT&T can bring high-speed internet to those same customers," says Rayburn. AT&T's Stephenson says that he expects the deal will mean 15 million DirecTV customers, many in rural areas, would be given access to broadband internet using a combination of technologies, including wireless. (DirecTV currently offers its customers satellite internet through packages with Exede, which can deliver download speeds of 12Mbps.)

In a world where double- and triple-play packages are virtually ubiquitous, this kind of upgrade will help ward off competition. "The one big thing we’ve been missing is a two-way broadband pipe to the home," says DirecTV CEO Mike White. "With this deal we can bundle video and broadband to combat the dominance of cable." Of course, the cry to combat "the dominance of cable" overlooks what White and Stephenson would prefer to obfuscate: DirecTV already counts over 20 million subscribers in the US alone, compared to 26.8 million for Comcast. By that metric, the satellite provider is no underdog.

DirecTV also gives AT&T a boost to its bottom line — it reported free cash flow around $2.6 billion last year — and gives the telecom giant substantial new inroads in Latin America: DirecTV currently counts more than 18 million customers there. On the call this morning, Stephenson made several mentions of AT&T's desire to expand aggressively in Latin America, especially Brazil.

The deal, like Comcast–Time Warner Cable, will need to pass regulatory scrutiny before it's approved. If both the mega-mergers do succeed, the media and communications landscape in America will have shifted dramatically. "You're talking about a world where two companies control 60 percent or more of the paid TV and internet market," says Rayburn. "If you thought the net neutrality folks were angry before, this is really gonna set them off."

Driverless Cars, Telematics and Actuarial Science

As reported by Insurance Networking News: Fewer accidents translate into lower insurance rates, and driverless car technologies could lower rates by as much as $475 per year, according to industry experts.

Driverless cars are on the streets in California, Nevada, Michigan and elsewhere, and while they are proving to cause fewer accidents than human drivers, there will continue to be accidents and someone will be liable for them. However, who will be responsible, and who will need insurance, is not yet clear.

At the Casualty Actuarial Society’s Ratemaking and Product Management seminar last month, industry experts discussed the implications of driverless cars and insurance telematics in the presentation “Autonomous Vehicles and the Impact on the Insurance Industry.” Presenters included Robert Peterson, a professor of law at Santa Clara University, Frank Douma, research fellow from the University of Minnesota, and Michael Stienstra, FCAS, AVP at QBE North America.

Human error contributes to more than 90 percent of all auto accidents, but even as accidents decline with consumer acceptance of the technology, there will continue to be a need to insurance the cars, their owners and manufacturers, the experts said.


Fewer accidents likely will means cheaper auto insurance, Peterson said, adding that rates could decline as much as $475 per year for operators of self-driving cars cheaper every year, according to a study by Alain L. Kornhauser, professor of operations research and financial engineering director, Transportation Program at Princeton University.

Many automakers say they will market driverless cars by 2020, and Google says it will have a fully automated car by 2017. Further, the costs may be lower than consumers may have imagined. According to Raj Rajkumar, director of the Carnegie Mellon-GM Autonomous Driving Collaborative Research Lab, quoted in the discussion, an autonomous technology package could add $5,000 to $7,000 to the sticker price.

For actuaries, the flood of data coming from driverless cars could be problematic. Stienstra said driverless cars likely will transmit as much as 750 megabytes of data per minute, and actuaries will have to cull the data before collecting it, and then find the variables that predict accidents.

Regulators also could create hurdles, Peterson said, adding that California has mandatory rating factors, including driving record and number of years as a driver; and that safer drivers receive discounts. With an automated vehicle, Peterson said those factors may prove irrelevant, and that state insurance laws will likely need to be altered to accommodate driverless cars.

Driverless cars may be safer than traditional cars, but flawed hardware or software could cause accidents, and liability could then fall on manufacturers or installers, in which case, the insurance pricing would fall to product liability actuaries for coverage.

For a considerable time, there will be a mixture of three types of cars, self-driving cars; partially automated cars, where the owner does some or almost all the driving, and human driven cars, Douma said. He described the five levels of vehicle automation, ranging from no automation, level 0, to fully self-driving, level 4. Currently, development efforts are aimed at level 3, where cars perform all safety-critical functions under certain conditions. Drivers will be alternating with computers for some time, Douma explains, as drivers could back out of the garage and onto the street before handing control to the computer, which could hand control back to the driver as they approach their destination.

Regulators, automakers and the public will expect safer cars to translate into lower insurance premiums, Stienstra said, and actuaries will need to be proactive on this issue, noting that the Casualty Actuarial Society has an Automated Vehicles Task Force to make sure casualty actuaries have the ability to partner with engineers and researchers to properly understand and insure the risks.

Telematics
The increasing availability and consumer acceptance of insurance telematics, also called usage-based insurance also has implications for insurers and actuaries.

In separate discussions, Jim Weiss, FCAS for ISO, Jerel Cestkowski of American Family Insurance and Allen Greenberg, a senior policy analyst at the U.S. Department of Transportation, addressed telematics technology and implications for insurers and consumers.


When drivers are aware that their behavior is being monitored, they may drive more safely, Weiss said, adding that studies have found crash rates fell between 20 and 30 percent in cars monitored by telematics devices.

Monitoring is not cheap, though. Weiss said telematics devices currently cost about $100, last three years, and wireless communications for each device costs about $5 a month.

Urban Engines Launches to Use Data to Make City Transportation Better

As reported by GigaOm: Stanford computer science professor Balaji Prabhakar first became interested in how transportation systems move when, years ago, he got stuck in “the mother of all traffic jams” in India. Now, after two years in stealthy development, Prabhakar and his co-founder, former Google exec Shiva Shivakumar, are launching a startup called Urban Engines that is using data, algorithms and behavioral economics to help make cities less congested and urban transportation operate more efficiently.

In an office in downtown San Francisco this week, six stories above the blaring horns of buses and cars running up and down Market Street, Shivakumar and Prabhakar showed me a screen of a train system that could be any big city in the world — Sao Paulo, San Francisco, Bangalore. Prabhakar clicked the play button and we watched a geometrical visualization of the flow of train commuters moving into stations, getting on trains and getting off at their stops. Some trains were too full, some not full enough. It was mesmerizing to watch, in a weird way.

But for city planners and transportation operators it could be an entirely new way of doing things. Urban Engines has built a system that can take data mostly from commuter transit cards — the bus card, the subway card (if you’re in San Francisco, your Clipper card) — and use algorithms to infer information about how commuters and the transportation flow are behaving.

Without embedding any sensors in the subway or video cameras watching the platforms, Urban Engines can tell things like how long commuters were waiting, how many trains went by that were so full commuters couldn't get on and what the volume of each train car was throughout the day. It only needs the data from when the commuter enters and exits the station, and by knowing the aggregate of all the commuter data at the same time, it can infer how the system is operating. It can do the same thing with a city bus system.

Essentially, Urban Engines is taking the smallest and cleanest amount of data possible to map out the entire public transportation network. The approach means that it can build such a monitoring system much more inexpensively than comparable transportation systems that use sensors, video cameras or even people manually observing and counting. A surprising amount of city accountability around transportation comes from city workers standing next to potential problem areas and observing — not exactly efficient or accurate.

The second piece of Urban Engine’s idea comes from behavioral economics: offering incentives or punishments to shape behavior. After identifying problem spots in transportation systems, it can help city planners use incentives to make the systems run better. For example, if too many people are using buses early in the morning, incentives (like being entered into a lottery) could be provided to encourage commuters to take the bus an hour later. Or if some train stations are being over-used and others are being under-used, incentives could be provided to get more commuters to the under-used stations.

Many of the ideas around incentives are based on the work that Prabhakar was doing at Stanford. He’s the director of the Stanford Center for Societal Networks, which works on making “societal networks smarter, more scalable and more efficient.” After leaving Google, Shivakumar (who’s Urban Engines’ CEO) came to Stanford as a visiting researcher to reboot, and the pair have been working together ever since.

Even though this might be the first time you’re hearing about them, Urban Engines actually already has some early traction and a hot list of investors. It’s working with the World Bank to implement its system for the buses in Sao Paulo. It’s helping Singapore shift its train commuters from peak hours to off-peak hours. And it’s got an early deployment with the train system in Washington, D.C. It’s also done pilots projects in Bangalore, with Infosys, on the Stanford campus.

Last year, Urban Engines raised a Series A round from some of the most prominent investors in the Valley, including Andreessen Horowitz, Google Ventures, Eric Schmidt, Greylock, SV Angel and angel investor Ram Shriram. The executives didn’t disclose the size of the round, but they’re employing about 20 people.

The underlying macro trend behind Urban Engines is that the population is growing rapidly, and by 2050 there will be 9 billion people on the planet. Much of the growth is happening in cities, and worldwide, more people now live in cities than outside of cities. City transportation systems will only get increasingly crunched over the coming decades. City planners and transportation builders will need new tools to help manage the influx.