If you want to use one of today's major VR headsets, whether the Oculus Rift, the HTC Vive, or the PS VR, you have to accept the fact that there will be an illusion-shattering cable that tethers you to the small supercomputer that's powering your virtual world.
But researchers from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) may have a solution in MoVr, a wireless virtual reality system. Instead of using Wi-Fi or Bluetooth to transmit data, the research team’s MoVR system uses high-frequency millimeter wave radio to stream data from a computer to a headset wirelessly at dramatically faster speeds than traditional technology.
There have been a variety of approaches to solving this problem already. Smartphone-based headsets such as Google's Daydream View and Samsung's Gear VR allow for untethered VR by simply offloading the computational work directly to a phone inside the headset. Or the entire idea of VR backpacks, which allow for a more mobile VR experience by building a computer that's more easily carried. But there are still a lot of limitations to either of these solutions.
Latency is the whole reason a wireless solution hasn't worked so far. VR is especially latency-sensitive, along with the huge bandwidth requirements that VR needs to display the level of high-resolution video required for virtual reality to work. But the MIT team claims that the millimeter wave signals can transmit fast enough to make a wireless VR headset feasible.
The issue with using millimeter wave technology is that the signal needs a direct line of sight, and fares poorly when it encounters any obstacles. MoVR gets around this by working as a programmable mirror that can direct the direction of the signal to the headset even while it’s moving to always make sure the signal is transmitting directly to the headset's receivers.
For now, the MoVR is simply a prototype, with the team hoping to further shrink down the system to allow for multiple wireless headsets in one room without encountering signal interference. But even as a proof-of-concept, it's an interesting perspective on how virtual reality could one day work.
As reported by Backchannel: It could be the plot of an Ocean’s Eleven movie: A team of daring engineers heads to the wilds of Nevada to create a dazzling spectacle, defy officials, and walk off with a $680 million jackpot.
In May of this year, Otto unveiled a video of an 18-wheeler thundering down a freeway with no one at the wheel. The San Francisco-based startup had retrofitted a freight truck with lidar, video cameras, radar, and other sensors, enabling it to operate “completely driverless without a driver in the backseat,” according to co-founder Lior Ron at the time.
Even Google, which had been developing self-driving cars for eight years, still insists on having a human in the driving seat should anything go wrong. That a tiny startup emerging from stealth mode could confidently demonstrate an autonomous big rig on a public interstate with no one in the front seats suggested Otto’s technology was really special.
Within just a couple of months, ride-sharing unicorn Uber had acquired the fledgling Otto in exchange for 1 percent of the company’s stock (representing $680 million) and 20 percent of any future self-driving trucking profits — a payout that could exceed a billion dollars for Otto’s founders.
So far, just another Silicon Valley fairytale. Yet Otto’s rapid ascent involved a high-stakes gamble that could have landed the startup in legal hot water, rather than the arms of the world’s biggest transportation network company.
When Otto performed its test drive — the one shown in the May video — it did so despite a clear warning from Nevada’s Department of Motor Vehicles (DMV) that it would be violating the state’s autonomous vehicle regulations. When the DMV realized that Otto had gone ahead anyway, one official called the drive “illegal” and even threatened to shut down the agency’s autonomous vehicle program.
The story of why Otto flouted Nevada’s directive, how it got away with it, and what this means for the future of self-driving regulation stretches back to the dawn of driverless cars.
Otto co-founder Anthony Levandowski
On May 1, 2012, the world’s first government-run self-driving testtook place on Las Vegas’s famous Strip. In the driver’s seat of a heavily modified 2008 Toyota Prius was Chris Urmson, then the head of Google’s self-driving car project. Next to him sat Anthony Levandowski, the brilliant engineer whose startup 510 Systems had been acquired by Google the previous year, after it had built the tech giant’s first autonomous vehicles.
As the Prius nosed its way past casinos and tourists, its performance was being watched from the back seat by a driving examiner and by Bruce Breslow, then the head of the Nevada DMV. The test was a quick zip around Las Vegas’s surface streets and freeways on a route chosen by Google itself. Despite Urmson having to take control of the vehicle on two occasions, the Prius passed the test with flying colors.
That was hardly surprising, as Nevada’s first autonomous vehicle rules had been proposed and partly shaped by Levandowski himself, with the help of David Goldwater, a lobbyist for Google. The company could now test its self-driving cars anywhere in the state. It was a publicity coup for Nevada, and the first official validation of Google’s vision to replace fallible human drivers with flawless robots.
Over the next four years, however, Google never actually tested its driverless cars in Nevada. And despite welcoming Daimler’s semi-autonomous trucks to the state with a glitzy ceremony in 2015, Nevada’s reputation as a pioneer in autonomous driving ebbed away.
Otto co-founder Lior Ron
The year 2016 brought some big changes: The energetic Levandowski decided to strike out on his own. “I’m excited about bringing robots into the market, about having the most effect in the world,” he told me. He gathered a handful of colleagues from Google, including Lior Ron, and founded Otto, a startup dedicated to “reimagining transportation—not just improving it.”
Initially, that meant a fully self-driving taxi rather than driverless trucks. In April of this year, Steve Hill, executive director of the Nevada Governor’s Office of Economic Development (GOED), wrote an email to Jude Hurin, the administrator in charge of autonomous vehicle (AV) regulation at the DMV. Backchannel obtained access to this and many more emails and documents under public records legislation.
“Just wanted to let you know that a company is interested in operating completely autonomously, shuttling passengers on a fixed route in Vegas somewhere around the end of this year,” Hill wrote. “David Goldwater represents them and will be contacting your office soon. The company is headed by Anthony Levandowski. All this is confidential.”
The company was Otto, and its plan was to offer driverless rides from Las Vegas’s McCarran airport to the heart of the Strip. Breslow, now director of the state’s Department of Business and Industry, chimed in. “Sounds like the Google project Anthony has wanted to set up for the last four years,” he wrote to Hurin on April 29. “[…] Should be relatively simple to deploy but political backlash from cab industry will occur.”
Photo credit: Otto
Less than two weeks later, Goldwater announced a surprising shift in direction. “Otto will have their company launch on Tuesday of next week,” he wrote to Hill on May 12. “[…] They are filming some video of self-driving trucks in Nevada tomorrow for the launch on Tuesday (there is someone in the truck [with] a kill switch.) The filming is being done up in Reno.”
Hurin’s response to Goldwater was swift. “I am concerned….Every company we have dealt with in the past has obtained a test license in our state prior to the media event….Your video is going to produce questions directly to your company and Nevada DMV as to why they do not have a red test plate on the vehicle. I am also going to have Daimler Trucking and others calling me to ask why Otto is allowed to do this when they couldn’t.”
Nevada regulations are clear on this point. Before an autonomous vehicle can be operated on the state’s roads, it must be issued a testing license and special red license plates. It has to be able to capture driving data in case of crashes, have switches to engage and disengage the autonomous systems, and have a way to alert the human operator if it fails.
To obtain a license, Otto would have had to produce evidence of 10,000 miles of previous autonomous operation and submit a truck for a self-driving test, such as the one completed by Google in 2012. It would also need to post a $5 million bond and file reams of paperwork. Even with all those requirements fulfilled, Otto’s demo would need two people seated up front, one of them poised to take over in the event of a failure.
“Without that license you do not have authorization to drive this on our highway in a semi or fully autonomous mode,” wrote Hurin to Goldwater. “Having just a ‘kill switch’ is only part of the safety requirements…but regardless a license is required.”
Goldwater had a different take on the issue, replying, “The truck does have a DOT [Department of Transportation] number, but is not part of a specific trucking company. Therefore, I think the easy answer is it’s not a ‘test truck’ yet. The efforts of Nevada’s leaders are to try to get Otto and similar companies to move here. Most certainly, when the trucks are testing, all laws and regulations will be followed.”
But Hurin was adamant. In an email sent early the next morning, May 13, he warned Goldwater: “It does not matter if you have another State’s license plate or US DOT sticker on it…if they show the vehicle engaging the technology and it implies or is obvious that Nevada is the site that they have engaged the system for the video, then it is a violation of our current AV testing laws.”
In an email to Hurin, Breslow agreed: “I told him [Goldwater] the same thing you did and that he couldn’t just do this.” When Hill, the GOED official, asked whether Otto might be quickly issued a testing permit, Hurin pointed out the impossibility of achieving that literally overnight.
Otto’s founders were faced with a stark choice. They could submit to the DMV and undertake the laborious process of modifying, testing, and licensing their truck. This would likely take a month or more, and could risk their first-mover advantage in driverless trucking. Or the engineers could continue with their test as planned.
When I asked Ron where and how the truck had been tested, he said, “We’ve driven a couple of miles completely driverless without a driver in the backseat — not in California, in some other state. We did that on Saturday and will have video assets to show you later today.” Otto’s launch post on Medium confirmed this, saying, “We recently completed an autonomous demo of the technology on a public highway.”
Photo credit: Otto
When I watched Otto’s video, I could see its truck, without anyone in the front seats, passing roadside signs on Interstate 80 near Reno. I included that information in my story but wondered how Otto had obtained permission to operate in the state without a license. I subsequently filed a public records request to find out more.
In the meantime, Otto’s launch went smoothly. Terri Albertson, director of the Nevada DMV, noticed the Nevada reference in Backchannel’s story and suggested to Hurin that the department post it on their website. His reply was nothing short of explosive: “Nope!!…They have no license, they are testing no matter what anyone tries to tell me and they have no passengers in the vehicle when current law requires two people….This could very well damage Nevada’s Autonomous ‘Pioneer’ image and our relationships with current OEMs and automated companies who already have a license. Otto is driving these vehicles illegally and without the required $5 million bond that is needed to protect our citizens.”
Hurin even suggested the DMV might walk away from autonomous vehicles altogether. “If push comes to shove we may need to propose that GOED…transfer the AV testing and consumer responsibilities to them and no longer have this under the DMV,” he wrote to Albertson. “I will continue to defend this program and its integrity but if the pressures increase, then it needs to be placed under GOED.” GOED is a pro-growth department focused on attracting and stimulating new business in the state. It has not typically been tasked with overseeing or regulating the industries it serves. But the move did not happen, and no legal action was taken against Otto.
Since then, Otto’s trucks have carried a commercial cargo of beer in Colorado and are due to start testing this week in Ohio, all with the cooperation of state officials.
When I contacted Otto for comment, the company declined to be quoted. It walked back from its claims at the Nevada launch, when it had described its truck as operating “completely driverless without a safety driver in the backseat,” and its drive as “an autonomous demo.” Otto now says its truck did in fact have someone in the cab at the time. Otto also says that the self-driving truck is not an autonomous vehicle at all, at least on Nevada’s roads.
Nevada’s regulations define autonomous technology as “technology which is installed on a motor vehicle and which has the capability to drive the motor vehicle without the active control or monitoring of a human operator.” Because Otto’s truck had an engineer watching from further back in the cab, says the company, it does not count as an autonomous vehicle and thus does not need a permit.
Otto’s argument might not hold up if challenged, says Ryan Calo, a law professor who teaches a class on robotics law and policy at the University of Washington: “One question is whether or not monitoring counts if you’re not in the driver’s seat. Often customs wind up informing the law and the custom here is that other testers, like Google and Tesla, actually have a person sitting in front of the steering wheel.”
Bryant Walker Smith, a professor at the University of South Carolina, points out another difficulty: “If the autonomous law does not apply, what does? If the driver is using a laptop or cellphone, that might run afoul of driver distraction laws. And the highway patrol…could very readily conclude that a big rig without someone at the steering wheel is reckless.”
The DMV’s position remains that Otto’s truck is an autonomous vehicle under Nevada regulations, and that the company violated them. But it doesn’t really matter either way. Bizarrely, the autonomous driving regulations come with no penalties for breaking them.
“The autonomous regulations and policies are living, breathing pieces of this cutting-edge technology,” Hurin told me. “The Department will be submitting another set of AV regulatory amendments in 2017. One of the areas we have already earmarked for these regulations is penalties for this type of violation.”
Over the summer, Hurin kept pressuring Otto to apply for a testing license. Finally, he wrote directly to Levandowski: “We need to find out a specific time when you will be submitting your AV Test Application packet….It has been conveyed to me many times in the past that you are working on this, but to date have received nothing…We look forward to our partnership with Otto, which, as you would agree, must be built on honesty and integrity.”
Photo credit: Otto
But Otto was about to pivot once more. The company no longer wanted to test its trucks in Nevada, as it had previously indicated. Instead, in July it applied for a license to operate the state’s first Autonomous Technology Certification Facility (ATCF).
Before any self-driving tech can be sold for use in Nevada, such a facility must issue a certificate stating that it complies with safety requirements, including capturing data prior to collisions, complying with traffic laws, and having the right controls and indicators. The idea is that third-party organizations will take pressure off the state to assess autonomous technologies, providing a smooth route to market.
While Otto was in the process of putting its application together, the company was snapped up by Uber. Levandowski became Uber’s head of self-driving technologies, although Otto would continue to lead the ATCF program in Nevada. Otto had now come full circle, from a tiny startup intent on disrupting taxis in Las Vegas, to maverick autonomous truckers, to a key part of the biggest ride-sharing company on the planet.
The testing facility, a wholly owned subsidiary called Nevada ATCF LLC, received its license from the DMV at the end of September, and is currently hiring staff and remodeling a building in southwest Las Vegas. Otto says that the ATCF will only be used to certify vehicles developed by Otto and Uber.
It is another peculiarity of Nevada’s regulations that a company is not required to test its autonomous vehicles in the state prior to deploying them. The moment Uber licenses its own self-driving trucks or taxis at the ATCF, the vehicles will be ready for public use. “When they decide to sell their technology, they will already be licensed to do so,” Hurin said.
Not making an issue of Otto’s unlicensed demo certainly helped Nevada in the short term. It avoided alienating a leading player in driverless technology and probably helped secure its first autonomous testing facility. But this also came at a cost. Until Nevada clarifies its autonomous vehicle laws and introduces penalties for breaking them, it is difficult to see why any self-driving company should bother with the hassle and expense of certification.
The crumbling of Nevada’s regulatory credibility leaves only California with tough, consumer-focused laws around self-driving cars. In much of the rest of the US, autonomous driving is a free-for-all. Uber taxis are transporting passengers in Pittsburgh, Google’s self-driving prototypes are criss-crossing Texas, and Tesla’s cars are taking over the wheel nationwide, with little official testing or licensing of the technology beforehand.
That might accelerate innovation today, but it means there will be few regulatory tools in place to cope with the many ethical, logistical, and safety challenges that lie further down the self-driving road. And if industry experts can decide to simply skirt the requests of state regulators, the prospects for future regulation look dim.
Otto was nothing less than the perfect Vegas spectacle, equal parts technology, audacity, and showmanship, culminating in an overnight wedding to Uber. As Otto’s founders count their whirlwind winnings, Nevada has shown that well-meaning attempts at keeping tabs on new technology can evaporate just as fast.
As reported by Engadget: Sure, Tesla's first demo of full self-driving features was intriguing. But did you wonder what it was like from the car's point of view? You're about to find out. Tesla has posted another demo video that shows what an autonomous EV sees as it navigates local roads. As the clip illustrates, the cameras and sensors have to detect many, many different objects at any given moment: road lines, signs, lights, pedestrians and cars are among the many examples. The car even captures the "motion flow" of the environment to get a sense of where it's going. This isn't a flawless test run. You see the vehicle slow down unnecessarily for people running by the roadside (about 1:30 in), for example. It's also uncertain how well this sort of demo would work in less than favorable conditions, such as snowy weather or nighttime drives on poorly-lit roads. However, it's still impressive. The car is sufficiently aware of its environment that the driver doesn't have to touch the wheel once, even in relatively complex situations.
As reported by Seeker: Investigators probing why a SpaceX Falcon 9 rocket burst into flames on its launch pad two months ago have found the smoking gun. They discovered that the rocket's liquid oxygen accidentally got so cold it became solid. That transformation, in turn, triggered a chemical reaction with a carbon composite container holding liquid helium that is located inside the oxygen tank.
The helium vessel's failure set off a series of explosions that destroyed the rocket and its payload, a $200 million communications satellite that it was due to carry into orbit two days later. The accident occurred as the rocket was being fueled for a routine, pre-launch engine test on Sept. 1.
SpaceX chief Elon Musk called the problem surprising and said it had never been encountered in the history of rocketry.
But it's not completely unexpected. A panel of highly respected aerospace experts told NASA eight months before the accident that the lack of a re-circulation pump aboard the Falcon rocket could be a safety problem.
"We are concerned that there may be insufficient pre-cooling of the tank and plumbing with (SpaceX's) current planned oxidizer fill scenario," former astronaut Thomas Stafford and members of the International Space Station Advisory Committee wrote in a December 2015 letter to NASA.
The letter was released on Friday.
"Without re-circulation there may be stratification of oxidizer temperature that will cause a variation in the input conditions to the oxidizer pump," the letter said.
NASA's space shuttles, for example, used a re-circulation system and pressure lines to help keep liquid oxygen temperatures consistent, Mike Leinbach, former shuttle launch director, told Seeker.
"That's key," said Leinbach.
NASA also has a bit of history with liquid oxygen behaving badly with a composite fuel tank, which was developed as part of the experimental X-33 spaceship in the 1990s.
"The composite tank experienced cracks when fueled in development tests. Composite was used to save weight. Never could overcome it technically, and contributed a lot to the program demise," Leinbach wrote in an email.
Camera view of the inside of the second stage's LOX tank with Helium COPVs.
The Stafford letter raised the lack of a re-circulation pump as part of a broader safety concern about SpaceX's plan to fuel its rockets with astronauts aboard. All previous rockets used to launch humans have been fueled before a crew arrived at the launch pad.
SpaceX is one of two companies NASA has hired to fly astronauts to the space station beginning in 2018.
During a committee meeting last week, Stafford also questioned SpaceX's unorthodox practice of putting helium canisters inside the oxygen tank. The helium, which is part of pressurization system, typically is located outside of the oxygen tank.
SpaceX has not yet said what it will do to fix the fueling system issue, but Musk said Friday said he is aiming to return the Falcon 9 to flight in mid-December.
As reported by Engadget: Researchers can now probe connected devices, computers and cars for security vulnerabilities without risking a lawsuit. Last Friday, the FTC authorized changes to the Digital Millennium Copyright Act (DMCA) that will allow Americans to do hack their own electronic devices. Researchers can lawfully reverse engineer products and consumers can repair their vehicle's electronics, but the FTC is only allowing the exemptions for a two-year trial run. The FTC and US Library of Congress enacted similar legislation in 2014 that allows you to unlock your own smartphone. Until today, however, it was illegal to mess with the programs in your car, thermostat or tractor, thanks to strict provisions in the DMCA's Section 1201. That applied even to researchers probing the device security for flaws, a service that helps both the public and manufacturers. For example, researchers commandeered a Jeep on the road to show it could be done, an act that was technically illegal. You could have also been sued just for trying to repair your own electronics. In a well-publicized example, John Deere told farmers that they have no right to root around in the software that runs their tractor even when they're just trying to fix the damned thing. That issue alone prompted over 40,000 public comments to the US Copyright Office demanding stronger ownership rights.
DMCA 1201, and the rulemaking process, create unconstitutional restraints on speech, and need to be struck down by a court or fixed by Congress.
The exemptions have certain restrictions -- consumers are only allowed to do "good-faith" hacking on "lawfully-acquired" devices. That means, for instance, that you can still get in trouble if you gain unauthorized access to a device you don't own. Also, researchers can't probe internet services or public services like airlines either, meaning that the jet hack done last year would still be illegal now.
Groups like the Electronic Frontier Foundation, iFixit and Repair.org fought to have research and repair activities exempted from the DMCA, since they actually have nothing to do with copyright law. "You could be sued or even jailed for trying to understand the software in your devices, or for helping others do the same," the EFF wrote.
The new exemptions are nice, but critics are still fuming over the fact that they took a year to kick in and are only good for two years. Repair and research advocates say that the process for changing copyright law is unnecessarily expensive and onerous, too. "The one year delay ... was not only a violation of law, not only pointless, but actively counterproductive," the EFF wrote. "DMCA 1201, and the rulemaking process, create unconstitutional restraints on speech and need to be struck down by a court or fixed by Congress."
As reported by Bloomberg: On Friday evening as the sun descended over the old Hollywood set of “Desperate Housewives,” Elon Musk took to a stage and fired up his presentation about climate change. It was a strange scene, with hundreds of people crowded into the middle of a subtly artificial suburban neighborhood.
It wasn’t until about a minute into the speech that Musk casually let the crowd in on Tesla’s big secret. “The interesting thing is that the houses you see around you are all solar houses,” Musk said. “Did you notice?”
The answer, in short, was no. Like everyone else, I knew we were there to see Musk’s new “solar roof,” whatever that was supposed to mean. But try as I could as we walked in, I didn’t see anything that looked like it could carry an electric current. If anything, the slate and Spanish clay roofs looked a bit too nice for a television set. This is the future of solar, Musk proclaimed. “You’ll want to call your neighbors over and say ‘check out the sweet roof.’ It’s not a phrase you hear often.”
The roof tiles are actually made of textured glass. From most viewing angles, they look just like ordinary shingles, but they allow light to pass through from above onto a standard flat solar cell. The plan is for Panasonic to produce the solar cells and for Tesla to put together the glass tiles and everything that goes along with them. That’s all predicated on shareholders approving the $2.2 billion acquisition of SolarCity, the biggest U.S. rooftop installer, on Nov. 17.
Tesla says the tempered glass is “tough as steel,” and can weather a lifetime of abuse from the elements. It can also be fitted with heating elements to melt snow in colder climates. “It’s never going to wear out,” Musk said, “It’s made of quartz. It has a quasi-infinite lifetime.”
In a Q&A with reporters after the presentation, Musk said the tiles are comparable to competing high-efficiency solar panels. The current prototypes that Tesla engineers are working with reduce the efficiency of the underlying solar cell by just 2 percent. With further refinement, Musk said he hopes the microscopic louvers responsible for making the tiles appear opaque can be used to actually boost the efficiency of standard photovoltaic cells.
Putting the pieces together
The vision presented at Universal Studios in Los Angeles is the grand unification of Musk’s clean-energy ambitions. The audience was able to step into a future powered entirely by Tesla: a house topped with sculpted Tuscan solar tiles, where night-time electricity is stored in two sleek wall-hung Powerwall batteries, and where a Model 3 prototype electric car sits parked out front within reach of the home’s car charger.
Attracting less attention on Wisteria Lane was Tesla’s Powerwall 2, a major upgrade of its home battery for electricity storage. When the original Powerwall was released last year, I was skeptical. Mostly, it was just too pricey for the amount of power it provided, especially in the U.S. where electricity is cheap and most people can sell their excess solar power back to the grid. Version 2 is a much different product. It packs more than twice the capacity—14 kilowatt hours versus 6.4 kilowatt hours—for less than half the price after installation. 1 It includes a built-in Tesla-brand inverter and comes with a ten year, infinite-cycle warranty.
Electricity storage is crucial for future uptake of solar power. Already in some solar-heavy regions, more electricity is being produced during the middle of the day than people can consume, and utility prices spike in the evening hours when the sun goes down. In the U.S., some states are abandoning payments for daytime rooftop solar, undermining huge investments that families have made in their solar systems. The only recourse is for customers to use that electricity themselves, at night.
Like previous attempts at solar shingles, the solar-plus-battery package hasn’t really caught on yet. SolarCity’s total bundled sales thus far number in just the hundreds. But an argument can be made that the products just weren't compelling enough yet and the prices were still too high.
The Powerwall 2 may be the cheapest lithium ion battery for the home ever made when deliveries start in January. Tesla is selling the batteries at retail prices that are cheaper than the average manufacturing cost at most companies, according to data compiled by Bloomberg New Energy Finance. We "certainly expect it will move the market prices downwards as we saw last year with the first Powerwall," said Yayoi Sekine, a BNEF analyst who covers battery technology.
“The future is going to overwhelmingly be solar plus battery,” Musk said. "They go together like peanut butter and jelly."
Let’s wait and see
Powerwall 2 looks ready for primetime. The new solar shingles? Let’s wait until more details emerge. Tesla says we should expect a slow initial rollout beginning in about nine months. Within two years of production, the shingles could account for five percent of the five million roofs installed in the U.S. every year, said Peter Rive, SolarCity’s co-founder and chief technology officer. SolarCity, under the Tesla brand, would also continue to sell surface-mounted solar panels for homeowners who have no plans for replacing their existing roofs.
The pricing on the new solar roof is a bit—squishy. Musk said that someone who buys a Tesla roof will save money compared with someone who buys a comparable traditional roof plus electricity from the grid. But make no mistake: This will be a premium product, at least when it first rolls out. The terra cotta and slate roofs Tesla mimicked are among the most expensive roofing materials on the market. SolarCity CEO Lyndon Rive noted that the price of a conventional roof can vary widely, from $7,000 to $70,000—based on materials, size, complexity, location—so giving out firm prices of a solar roof at this point would be difficult.
Telsa will release more financial information about the SolarCity deal this week before it goes to shareholders for a vote. If all of Musk's plans come true, by the end of next year you'll be able to walk into a Tesla store, buy a Model 3 electric car, a slate-glass solar roof, and a Powerwall 2 to manage the flow of all those electrons in your life. There are a lot of details to be hammered out until we know for certain whether Musk’s vision for a grand unification will become more than just a great television backdrop. But these tiles, viewed up close, are definitely worth tuning in for.
Before it's here, it's on the Bloomberg Terminal.LEARN MORE
As reported by Wired: IN LESS THAN 12 hours, three different people offered to pay me if I’d spend an hour talking to a stranger on the phone.
All three said they’d enjoyed reading an article I’d written about Google building a new computer chip for artificial intelligence, and all three urged me to discuss the story with one of their clients. Each described this client as the manager of a major hedge fund, but wouldn’t say who it was.
The requests came from what are called expert networks—research firms that connect investors with people who can help them understand particular markets and provide a competitive edge (sometimes, it seems, through insider information). These expert networks wanted me to explain how Google’s AI processor would affect the chip market. But first, they wanted me to sign a non-disclosure agreement. I declined.
These unsolicited, extremely specific, high-pressure requests—which arrived about three week ago—underscore the radical changes underway in the enormously lucrative computer chip market, changes driven by the rise of artificial intelligence. Those hedge fund managers see these changes coming, but aren’t quite sure how they’ll play out.
Of course, no one is quite sure how they’ll play out.
Today, Internet giants like Google, Facebook, Microsoft, Amazon, and China’s Baidu are exploring a wide range of chip technologies that can drive AI forward, and the choices they make will shift the fortunes of chipmakers like Intel and nVidia. But at this point, even the computer scientists within those online giants don’t know what the future holds.
Going Deep
These companies run their online services from data centers packed with thousands of servers, each driven by a chip called a central processing unit, or CPU. But as they embrace a form of AI called deep neural networks, these companies are supplementing CPUs with other processors. Neural networks can learn tasks by analyzing vast amounts of data, including everything from identifing faces and objects in photos to translating between languages, and they require more than just CPU power.
And so Google built the Tensor Processing Unit, or TPU. Microsoft is using a processor called a field programmable gate array, or FPGA. Myriad companies employ machines equipped with vast numbers of graphics processing units, or GPUs. And they’re all looking at a new breed of chip that could accelerate AI from inside smartphones and other devices.
Any choice these companies make matters, because their online operations are so vast. They buy and operate far more computer hardware than anyone else on Earth, a gap that will only widen with the continued importance of cloud computing. If Google chooses one processor over another, it can fundamentally shift the chip industry.
The TPU poses a threat to companies like Intel and nVidia because Google makes this chip itself. But GPUs also play an enormous role within Google and its ilk, and nVidia is the primary manufacturer of these specialized chips. Meanwhile, Intel has inserted itself into the mix by acquiring Altera, the company that sells all those FPGAs to Microsoft. At $16.7 billion, it was Intel’s largest acquisition ever, which underscores just how much the chip market is changing.
First, Training. Then, Execution
But sorting all this out is difficult—in part because neutral networks operate in two stages. The first is the training stage, where a company like Google trains the neural network to perform a given task, like recognizing faces in photos or translating from one language to another. The second is the execution stage, where people like you and me actually use the neural net—where we, say, post a photo of our high school reunion to Facebook and it automatically tags everyone in it. These two stages are quite different, and each requires a different style of processing.
Today, GPUs are the best option for training. Chipmakers designed GPUs to render images for games and other highly graphical applications, but in recent years, companies like Google discovered these chips can also provide an energy-efficient means of juggling the mind-boggling array of calculations required to train a neural network. This means they can train more neural nets with less hardware. Microsoft AI researcher XD Huang calls GPUs “the real weapon.” Recently, his team completed a system that can recognize certain conversational speech as well as humans, and it took them about a year. Without GPUs, he says, it would have taken five. After Microsoft published a research paper on this system, he opened a bottle of champagne at the home of Jen-Hsun Huang, the CEO of nVidia.
But companies also need chips that can rapidly execute neural networks, a process called inference. Google built the TPU specifically for this. Microsoft uses FPGAs. And Baidu is using GPUs, which aren’t as well suited to inference as they are to training, but can do the job with the right software in place.
To the Smartphone
At the same time, others are building chips to help execute neural networks on smartphones and other devices. IBM is building such a chip, though some wonder how effective it might be. And Intel has agreed to acquire Movidius, a company that is already pushing chips into devices.
Intel understands that the market is changing. Four years ago, the chip maker told us it sells more server processors to Google than it sells to all but four other companies—so it sees firsthand how Google and its ilk can shift the chip market. As a result, it’s now placing bets everywhere. Beyond snapping up Altera and Movidius, it has agreed to buy a third AI chip company called Nervana.
That makes sense, because the market is only starting to develop. “We’re now at the precipice of the next big wave of growth,” Intel vice president Jason Waxman recently told me, “and that’s going to be driven by artificial intelligence.” The question is where the wave will take us.
An AI feedback loop annotates 'noise' with ghostly images it's been trained to see.