Porsche’s venture arm has acquired a minority stake in TriEye, an Israeli startup that’s working on a sensor technology to help vehicle driver-assistance and self-driving systems see better in poor weather conditions like dust, fog and rain.
The strategic investment is part of a Series A financing round that has been expanded to $19 million. The round was initially led by Intel Capital and Israeli venture fund Grove Ventures. Porsche has held shares in Grove Ventures since 2017.
TriEye has raised $22 million to date. Terms of Porsche’s investment were not disclosed.
The additional funding will be used for ongoing product development, operations and hiring talent, according to TriEye.
The advanced driver-assistance systems found in most new vehicles today typically rely on a combination of cameras and radar to “see.” Autonomous vehicle systems, which are being developed and tested by dozens of companies such as Argo AI, Aptiv, Aurora, Cruise and Waymo, have a more robust suite of sensors that include light detection and ranging radar (lidar) along with cameras and ultrasonic sensors.
For either of these systems to function properly, they need to be able to see in all conditions. This pursuit of sensor technology has sparked a boom in startups hoping to tap into demand from automakers and companies working on self-driving car systems.
TriEye is one of them. The premise of TriEye is to solve the low visibility problem created by poor weather conditions. The startup’s co-founders argue that fusing existing sensors such as radar, lidar and standard cameras don’t solve this problem.
TriEye, which was founded in 2017, believes the answer is through short-wave infrared (SWIR) sensors. The startup said it has developed an HD SWIR camera that is a smaller size, higher resolution and cheaper than other technologies. The camera is due to launch in 2020.
The technology is based on advanced nano-photonics research by Uriel Levy, a TriEye co-founder and CTO who is also a professor at the Hebrew University of Jerusalem.
The company says its secret sauce is its “unique” semiconductor design that will make it possible to manufacture SWIR HD cameras at a “fraction of their current cost.”
TriEye’s technology was apparently good enough to get Porsche’s attention.
Michael Steiner, a Porsche AG board member focused on R&D, said the technology was promising, as was the team, which is comprised of people with expertise in deep learning, nano-photonics and semiconductor components.
“We see great potential in this sensor technology that paves the way for the next generation of driver assistance systems and autonomous driving functions,” Steiner said in a statement. “SWIR can be a key element: it offers enhanced safety at a competitive price.”
Postmates has officially received the green light from the city of San Francisco to begin testing its Serve wheeled delivery robot on city streets, as first reported by the SF Chronicle and confirmed with Postmates by TechCrunch. The on-demand delivery company told us last week that it expected the issuance of the permit to come through shortly after a conditional approval, and that’s exactly what happened on Wednesday thes week.
The permit doesn’t cover the entire city – just a designated area of a number of blocks in and around Potrero Hill and the Inner Mission, but it will allow Postmates to begin testing up to three autonomous delivery robots at once, at speeds of up to 3 mph. Deliveries can only take place between 8 AM and 6:30 PM on weekdays, and a human has to be on hand within 30 feet of the vehicles while they’re operating at all times. Still, it’s a start, and green light for a city regulatory environment that has had a somewhat rocky start with some less collaborative early pilots from other companies.
Autonomous delivery bot company Marble also has a permit application pending with the city’s Public Works department, and will look to test its own four-wheeled, sensor-equipped rolling delivery bots within the city soon should it be granted similar testing approval.
Postmates first revealed Serve last December, taking a more anthropomorphic approach to the vehicle’s overall design. Like many short-distance delivery robots of its ilk, it includes a lockable cargo container and screen-based user interface for eventual autonomous deliveries to customers. The competitive field for autonomous rolling delivery bots is growing continuously, with companies like Starship Technologies, Amazon and many more throwing their hats in the ring.
UPS said Thursday it has taken a minority stake in self-driving truck startup TuSimple just months after the two companies began testing the use of autonomous trucks in Arizona.
The size of minority investment, which was made by the company’s venture arm UPS Ventures, was not disclosed. The investment and the testing comes as UPS looks for new ways to remain competitive, cut costs and boost its bottom line.
TuSimple, which launched in 2015 and has operations in San Diego and Tucson, Arizona, believes it can deliver. The startup says it can cut average purchased transportation costs by 30%.
TuSimple, which is backed by Nvidia, ZP Capital and Sina Corp., is working on a “full-stack solution,” a wonky industry term that means developing and bringing together all of the technological pieces required for autonomous driving. TuSimple is developing a Level 4 system, a designation by the SAE that means the vehicle takes over all of the driving in certain conditions.
An important piece of TuSimple’s approach is its camera-centric perception solution. TuSimple’s camera-based system has a vision range of 1,000 meters, the company says.
The days of when highways will be filled with autonomous trucks are years away. But UPS believes it’s worth jumping in at an early stage to take advantage of some of the automated driving such as advanced braking technology that TuSimple can offer today.
“UPS is committed to developing and deploying technologies that enable us to operate our global logistics network more efficiently,” Scott Price, chief strategy officer at UPS said in a statement. “While fully autonomous, driverless vehicles still have development and regulatory work ahead, we are excited by the advances in braking and other technologies that companies like TuSimple are mastering. All of these technologies offer significant safety and other benefits that will be realized long before the full vision of autonomous vehicles is brought to fruition — and UPS will be there, as a leader implementing these new technologies in our fleet.”
UPS initially tapped TuSimple to help it better understand how Level 4 autonomous trucking might function within its network. That relationship expanded in May when the companies began using self-driving tractor trailers to carry freight on a freight route between Tucson and Phoenix to test if service and efficiency in the UPS network can be improved. This testing is ongoing. All of TuSimple’s self-driving trucks operating in the U.S. have a safety driver and an engineer in the cab.
TuSimple and UPS monitor all aspects of these trips, including safety data, transport time and the distance and time the trucks travel autonomously, the companies said Thursday.
UPS isn’t the only company that TuSimple is hauling freight for as part of its testing. TuSimple has said its hauling loads for for several customers in Arizona. The startup has a post-money valuation of $1.095 billion (aka unicorn status).
We all see the headlines nearly every day. A drone disrupting the airspace in one of the world’s busiest airports, putting aircraft at risk (and inconveniencing hundreds of thousands of passengers) or attacks on critical infrastructure. Or a shooting in a place of worship, a school, a courthouse. Whether primitive (gunpowder) or cutting-edge (unmanned aerial vehicles) in the wrong hands, technology can empower bad actors and put our society at risk, creating a sense of helplessness and frustration.
Current approaches to protecting our public venues are not up to the task, and, frankly appear to meet Einstein’s definition of insanity: “doing the same thing over and over and expecting a different result.” It is time to look past traditional defense technologies and see if newer approaches can tilt the pendulum back in the defender’s favor. Artificial Intelligence (AI) can play a critical role here, helping to identify, classify and promulgate counteractions on potential threats faster than any security personnel.
Using technology to prevent violence, specifically by searching for concealed weapons has a long history. Alexander Graham Bell invented the first metal detector in 1881 in an unsuccessful attempt to locate the fatal slug as President James Garfield lay dying of an assassin’s bullet. The first commercial metal detectors were developed in the 1960s. Most of us are familiar with their use in airports, courthouses and other public venues to screen for guns, knives and bombs.
However, metal detectors are slow and full of false positives – they cannot distinguish between a Smith & Wesson and an iPhone. It is not enough to simply identify a piece of metal; it is critical to determine whether it is a threat. Thus, the physical security industry has developed newer approaches, including full-body scanners – which are now deployed on a limited basis. While effective to a point, the systems in use today all have significant drawbacks. One is speed. Full body scanners, for example, can process only about 250 people per hour, not much faster than a metal detector. While that might be okay for low volume courthouses, it’s a significant problem for larger venues like a sporting arena.
Image via Getty Images
Fortunately, new AI technologies are enabling major advances in physical security capabilities. These new systems not only deploy advanced sensors to screen for guns, knives and bombs, they get smarter with each screen, creating an increasingly large database of known and emerging threats while segmenting off alarms for common, non-threatening objects (keys, change, iPads, etc.)
As part of a new industrial revolution in physical security, engineers have developed a welcomed approach to expediting security screenings for threats through machine learning algorithms, facial recognition, and advanced millimeter wave and other RF sensors to non-intrusively screen people as they walk through scanning devices. It’s like walking through sensors at the door at Nordstrom, the opposite of the prison-like experience of metal detectors with which we are all too familiar. These systems produce an analysis of what someone may be carrying in about a hundredth of a second, far faster than full body scanners. What’s more, people do not need to empty their pockets during the process, further adding speed. Even so, these solutions can screen for firearms, explosives, suicide vests or belts at a rate of about 900 people per hour through one lane.
Using AI, advanced screening systems enable people to walk through quickly and provide an automated decision but without creating a bottleneck. This volume greatly improves traffic flow while also improving the accuracy of detection and makes this technology suitable for additional facilities such as stadiums and other public venues such as Lincoln Center in New York City and the Oakland airport.
Apollo Shield’s anti-drone system.
So much for the land, what about the air? Increasingly drones are being used as weapons. Famously, this was seen in a drone attack last year against Venezuelan president Nicolas Maduro. An airport drone incident drew widespread attention when a drone shut down Gatwick Airport in late 2018 inconveniency stranded tens of thousands of people.
People are rightly concerned about how easy it is to get a gun. Drones are also easy to acquire and operate, and quite difficult to monitor and to defend against. AI is now being deployed to prevent drone attacks, whether at airports, stadiums, or critical infrastructure. For example, new AI-powered radar technology is being used to detect, classify, monitor and safely capture drones identified as dangerous.
Additionally, these systems use can rapidly develop a map of the airspace and effectively create a security “dome” around specific venues or areas. These systems have an integration component to coordinate with on-the-ground security teams and first responders. Some even have a capture drone to incarcerate a suspicious drone. When a threatening drone is detected and classified by the system as dangerous, the capture drone is dispatched and nets the invading drone. The hunter then tows the targeted drone to a safe zone for the threat to be evaluated and if needed, destroyed.
While there is much dialogue about the potential risk of AI affecting our society, there is also a positive side to these technologies. Coupled with our best physical security approaches, AI can help prevent violent incidents.
In two years, Voyage has gone from a tiny self-driving car upstart spun out of Udacity to a company able to operate on 200 miles of roads in retirement communities.
Now, Voyage is on the verge of introducing a new vehicle that is critical to its mission of launching a truly driverless ride-hailing service. (Human safety drivers not included.)
This internal milestone, which Voyage CEO Oliver Cameron hinted at in a recent Medium post, went largely unnoticed. Voyage, after all, is just a 55-person speck of a startup in an industry, where the leading companies have amassed hundreds of engineers backed by war chests of $1 billion or more. Voyage has raised just $23.6 million from investors that include Khosla Ventures, CRV, Initialized Capital and the venture arm of Jaguar Land Rover.
Still, the die has yet to be cast in this burgeoning industry of autonomous vehicle technology. These are the middle-school years for autonomous vehicles — a time when size can be misinterpreted for maturity and change occurs in unpredictable bursts.
The upshot? It’s still unclear which companies will solve the technical and business puzzles of autonomous vehicles. There will be companies that successfully launch robotaxis and still fail to turn their service into a profitable commercial enterprise. And there will be operationally savvy companies that fail to develop and validate the technology to a point where human drivers can be removed.
Voyage wants to unlock both.
China’s EHang, a company focused on developing and deploying autonomous passenger and freight low-altitude vehicles, will build out its first operational network of air taxis and transports in Guangzhou. The company announced that the Chinese city would play host to its pilot location for a citywide deployment.
The pilot will focus on not only showing that a low-altitude, rotor-powered aircraft makes sense for use in cities, but that a whole network of them can operate autonomously in concert, controlled and monitored by a central traffic management hub that Ehang will develop together with the local Guangzhou government.
Ehang, which was chosen at the beginning of this year by China’s Civil Aviation Administration as the sole pilot company to be able to build out autonomous flying passenger vehicle services, has already demonstrated flights of its Ehang 184 vehicles carrying passengers in Vienna earlier this year, and ran a number of flights in Guangzhou in 2018 as well.
In addition to developing the air traffic control system to ensure that these operate safely as a fleet working in the air above city at the same time, Ehang will be working with Guangzhou to build out the infrastructure needed to operate the network. The plan for the pilot is to use the initial stages to continue to test out the vehicles, as well as the vertiports it’ll need to support their operation, and then it’ll work with commercial partners for good transportation first.
The benefits of such a network will be especially valuable for cities like Guangzhou, where rapid growth has led to plenty of traffic and high density at the ground level. It could also potentially have advantages over a network of autonomous cars or wheeled vehicles, since those still have to contend with ground traffic, pedestrians, cyclists and other vehicles in order to operate, while the low-altitude air above a city is more or less unoccupied.
A year after coming out of stealth mode with $40 million, self-driving truck startup Kodiak Robotics will begin making its first commercial deliveries in Texas.
Kodiak will open a new facility in North Texas to support it freight operations along with increased testing in the state. The commercial route
There are some caveats to the milestone. Kodiak’s self-driving trucks will have a human safety driver behind the wheel. And it’s unclear how significant this initial launch is; the company didn’t provide details on who its customers are or what it will be hauling.
Kodiak has eight autonomous trucks in its fleet, and according to the company it’s “growing quickly.”
Still, it does mark progress for such a young company, which co-founders Don Burnette and Paz Eshel say is due to its talented and experienced workforce.
Burnette, who is CEO of Kodiak, was part of the Google self-driving project before leaving and co-founding Otto in early 2016, along with Anthony Levandowski, Lior Ron and Claire Delaunay. Uber would acquire Otto (and its co-founders). Burnette left Uber to launch Kodiak in April 2018 with Eshel, a former venture capitalist and now the startup’s COO.
In August 2018, the company announced it had raised $40 million in Series A financing led by Battery Ventures . CRV, Lightspeed Venture Partners and Tusk Ventures also participated in the round. Itzik Parnafes, a general partner at Battery Ventures, joined Kodiak’s board.
Kodiak is the latest autonomous vehicle company to test its technology in Texas. The state has become a magnet for autonomous vehicle startups, particularly those working on self-driving trucks. That’s largely due to the combination of a friendly regulatory environment and the state’s position as a logistics and transportation hub.
“As a region adding more than 1 million new residents each decade, it is important to develop a comprehensive strategy for the safe and reliable movement of people and goods,” Thomas Bamonte, senior program manager of Automated Vehicles for the North Central Texas Council of Governments, said in a statement. “Our policy officials on the Regional Transportation Council have been very forward-thinking in their recognition of technology as part of the answer, which is positioning our region as a leader in the automated vehicle industry.”
Self-driving truck startup TuSimple was awarded a contract this spring to complete five round trips, for a two-week pilot, hauling USPS trailers more than 1,000 miles between the postal service’s Phoenix and Dallas distribution centers. A safety engineer and driver will be on board throughout the pilot.
Other companies developing autonomous vehicle technology for trucks such as Embark and Starsky Robotics have also tested on Texas roads.
Here at TechCrunch, we like to think about what’s next, and there are few technologies quite as exotic and futuristic as quantum computing. After what felt like decades of being “almost there,” we now have working quantum computers that are able to run basic algorithms, even if only for a very short time. As those times increase, we’ll slowly but surely get to the point where we can realize the full potential of quantum computing.
For our TechCrunch Sessions: Enterprise event in San Francisco on September 5, we’re bringing together some of the sharpest minds from some of the leading companies in quantum computing to talk about what this technology will mean for enterprises (p.s. early-bird ticket sales end this Friday). This could, after all, be one of those technologies where early movers will gain a massive advantage over their competitors. But how do you prepare yourself for this future today, while many aspects of quantum computing are still in development?
IBM’s quantum computer demonstrated at Disrupt SF 2018
Joining us onstage will be Microsoft’s Krysta Svore, who leads the company’s Quantum efforts; IBM’s Jay Gambetta, the principal theoretical scientist behind IBM’s quantum computing effort; and Jim Clark, the director of quantum hardware at Intel Labs.
That’s pretty much a Who’s Who of the current state of quantum computing, even though all of these companies are at different stages of their quantum journey. IBM already has working quantum computers, Intel has built a quantum processor and is investing heavily into the technology and Microsoft is trying a very different approach to the technology that may lead to a breakthrough in the long run but that is currently keeping it from having a working machine. In return, though, Microsoft has invested heavily into building the software tools for building quantum applications.
During the panel, we’ll discuss the current state of the industry, where quantum computing can already help enterprises today and what they can do to prepare for the future. The implications of this new technology also go well beyond faster computing (for some use cases); there are also the security issues that will arise once quantum computers become widely available and current encryption methodologies become easily breakable.
The early-bird ticket discount ends this Friday, August 9. Be sure to grab your tickets to get the max $100 savings before prices go up. If you’re a startup in the enterprise space, we still have some startup demo tables available! Each demo table comes with four tickets to the show and a high-visibility exhibit space to showcase your company to attendees — learn more here.
Self-driving startup Optimus Ride will become the first to operate a commercial self-driving service in the state of New York – in Brooklyn. But don’t expect these things to be contending with pedestrians, bike riders, taxis and cars on New York’s busiest roads; instead, they’ll be offering shuttle services within Brooklyn Navy Yards, a 300-acre private commercial development.
The Optimus Ride autonomous vehicles, which have six seats across three rows for passengers, and which also always have both a safety driver and another Optimus staff observer on board, at least for now, will offer service seven days a week, for free, running a service loop that will cover the entire complex. It includes a stop at a new ferry landing on-site, which means a lot of commuters should be able to pretty easily grab a seat in one for their last-mile needs.
Optimus Ride’s shuttles have been in operation in a number of different sites across the U.S., including in Boston, Virginia, California and Massachusetts.
The Brooklyn Navy Yards is a perfect environment for the service, since it plays host to some 10,000 workers, but also includes entirely private roads – which means Optimus Ride doesn’t need to worry about public road rules and regulations in deploying a commercial self-driving service.
May Mobility, an Ann Arbor-based startup also focused on low-speed autonomous shuttles, has deployed in partnership with some smaller cities and on defined bus route paths. The approach of both companies is similar, using relatively simple vehicle designs and serving low-volume ridership in areas where traffic and pedestrian patterns are relatively easy to anticipate.
Commercially viable, fully autonomous robotaxi service for dense urban areas is still a long, long way off – and definitely out of reach for startup and smaller companies in the near-term. Tackling commercial service in controlled environments on a smaller scale is a great way to build the business while bringing in revenue and offering actual value to paying customers at the same time.
Hyperloop, the futuristic and still theoretical transportation system that could someday propel people and packages at speeds of more than 600 miles per hour, has been designated a “public infrastructure project” by India lawmakers in the state of Maharashtra.
Wrapped in that government jargon is a valuable and notable outcome. The upshot: hyperloop is being treated like any other public infrastructure project such as bridges, roads and railways. In other words, hyperloop has been plucked out of niche, futuristic obscurity and given a government stamp of approval.
That’s remarkable, considering that the idea for hyperloop was first proposed by Tesla and SpaceX CEO Elon Musk in a nearly 60-page public white paper just five years ago.
It also kicks off a process that could bring hyperloop to a 93-mile stretch of India between the cities of Mumbai and Pune. The Pune Metropolitan Regional Development Authority will begin the procurement process in mid-August when it starts accepting proposals from companies hoping to land the hyperloop contract.
The frontrunner is likely Virgin Hyperloop One -DP World, a consortium between the hyperloop company and its biggest backer that pitched the original project to India. The MahaIDEA Committee earlier approved Virgin Hyperloop One-DP World Consortium as the Original Project Proponent.
Under the VHO-DPW proposal, a hyperloop capable of transporting 200 million people every year would be built between Pune and Mumbai. That stretch of road now takes more than three hours by car; VHO says its hyperloop would reduce it to a 35-minute trip.
“This is history in the making. The race is on to host the first hyperloop transportation system in the world, and today’s announcement puts India firmly in the lead. This is a significant milestone and the first of many important steps toward bringing hyperloop to the masses,” Virgin Hyperloop One CEO Jay Walder said in a statement Wednesday.
The hope is that India’s government will award the contract by the end of 2019, a VHO executive told TechCrunch. If that occurs, Phase 1 of the project — an 11.8 kilometer (or 7.3 mile) section — would begin in 2020.
The cost of building Phase 1 will be covered by DP World, which has committed $500 million to this section. The government is covering the cost and logistics of acquiring the land for the hyperloop.
Phase 1 will initially act as a certification track, which will be used to certify the hyperloop technology for passenger operations. VHO wants this certification track built and operating by 2024. If this section meets safety standards it will become part of the larger hyperloop line between Pune and Mumbai.
There is a lot of work to do, and technical milestones to meet, before hyperloop is whisking people in pods through a tunnel. But if it works and is built, the region’s economy could be transformed, supporters insist.
Once commercialized, the hyperloop will transform the Pune-Mumbai corridor into a mega-economic region, according to Harj Dhaliwal, managing director of India and Middle East at Virgin Hyperloop One.
Today, some 75 million people travel between Pune and Mumbai each year, and forecasts suggest that number could rise to 130 million annually by 2026. The VHO-DPW consortium says its hyperloop will have the capacity to handle 16,000 passengers day, or about 200 million people annually.
The AWS DeepRacer is an almost toylike 1/18th scale race car. It comes with all of the sensors and software tools to help developers build machine learning models to drive the car around a course — or really do anything else they want it to do. The $399 DeepRacer launched at AWS’s massive re:Invent show in late 2018.
At the time, it seemed like a bit of a gimmick, but AWS has put a lot of its weight behind it and is currently running a DeepRacer league at its various events around the world. At these events, developers can pit their models against each other and learn more about building a specific kind of machine learning model in the process.
Why bother, though? It’s not like DeepRacer cars are likely to add to AWS’s bottom line anytime soon. DeepRacer, however, is part of a line of hardware products from AWS that started with DeepLens, a smart camera for developers.
“It really comes from the same place,” AWS general manager for Artificial Intelligence and Machine Learning marketing Ryan Gavin told me. “When you think about the stimulus for something like DeepLens, it was really about how do we put machine learning into the hands of every developer and data scientist. That’s our mission and we’re very consistent about that.”
Ford has acquired a small robotics company based in Michigan called Quantum Signal, which has produced mobile robots for a number of clients, including the U.S. military. The company’s specialty has been building remote control software for robotic vehicles, specifically, and it’s also responsible for a very highly regarded simulated testing and development environment for autonomous and remotely controlled robotic systems.
All of the above is useful not only when developing military robots, but also when setting out to build and deploy self-driving cars — hence Ford’s interest in acquiring Quantum Signal. Ford said in a blog post that while others might’ve been sleeping on Quantum Signal and the work it has done, it has been following the company closely, and will employ its experience in developing real-time simulation and algorithms related to autonomous vehicle control systems to help build out Ford’s self-driving vehicles, transportation-as-a-service platform and hardware and software related to both.
Reading between the lines here, it sounds like Ford’s main interest was in picking up some experienced talent working on autonomy, and very specific challenges that are needed to develop road-worthy self-driving vehicles, including perception systems and virtual testing environments. Ford does, however, explicitly lay out a desire to “preserve” Quantum’s own “unique culture” as it brings the company on-board, pointing out that that’s the course it took with similar acquisition SAIPS (an Israeli computer vision and machine learning company) when it brought that team on-board in 2016.
SAIPS has now more than doubled its team to 30 people, and relocated to a new headquarters in Tel Aviv, with a specific focus among its latest hires on bringing in specialists in reinforcement learning. Ford has also invested in Argo AI, taking a majority stake in the startup initially in 2017 and then re-upping with a joint investment with Volkswagen in July of this year in a deal that makes both major equal shareholders. Ford is happy to both acquire and partner in its pursuit of self-driving tech development, and this probably won’t be the last similar deal we see made en route to actually deploying autonomous vehicles on roads for any major automaker.
The business, which is similar to startups in the U.S. like Filld, Yoshi and Booster Fuels, took 10 months to design and receive approval for its proprietary refueling trucks that can withstand the unique stresses of providing logistics services in India.
Together with co-founder Nabin Roy, a serial startup entrepreneur, MyPetrolPump co-founder and chief executive Ashish Gupta pooled $150,000 to build the company’s first two refuelers and launch the business.
MyPetrolPump began operating out of Bangalore in 2017 working with a manufacturing partner to make the 20-30 refuelers that the company expects it will need to roll out its initial services. However, demand is far outstripping supply, according to Gupta.
“We would need hundreds of them to fulfill the demand,” Gupta says. In fact the company is already developing a licensing strategy that would see it franchise out the construction of the refueling vehicles and regional management of the business across multiple geographies.
Bootstrapped until this $1.6 million financing, MyPetrolPump already has five refueling vehicles in its fleet and counts 2,000 customers already on its ledger.
These are companies like Amazon and Zoomcar, which both have massive fleets of vehicles that need refueling. Already the company has delivered 5 million liters of fuel with drivers working daily 12-hour shifts, Gupta says.
While services like MyPetrolPump have cropped up in the U.S. as a matter of convenience, in the Indian context, the company’s offering is more of necessity, says Gupta.
“In the Indian context, there’s pilferage of fuel,” says Gupta. Bus drivers collude with gas station operators to skim money off the top of the order, charging for 50 liters of fuel but only getting 40 liters pumped in. Another problem that Gupta says is common is the adulteration of fuel with additives that can degrade the engine of a vehicle.
There’s also the environmental benefit of not having to go all over to refill a vehicle, saving fuel costs by filling up multiple vehicles with a single trip from a refueling vehicle out to a location with a fleet of existing vehicles.
The company estimates it can offset 1 million tons of carbon in a year — and provide more than 300 billion liters of fuel. The model has taken off in other geographies as well. There’s Toplivo v Bak in Russia (which was acquired by Yandex), Gaston in Paris and Indonesia’s everything mobility company, Gojek, whose offerings also include refueling services.
And Gupta is preparing for the future as well. If the world moves to electrification and electric vehicles, the entrepreneur says his company can handle that transition as well.
“We are delivering a last-mile fuel delivery system,” says Gupta. “If tomorrow hydrogen becomes the dominant fuel we will do that… If there is electricity we will do that. What we are building is the convenience of last-mile delivery to energy at the doorstep.”
Alphabet’s autonomous driving and robotaxi company Waymo does a lot of training in order to refine and improve the artificial intelligence that powers its self-driving software. Recently, it teamed up with fellow Alphabet company and AI specialist DeepMind to develop new training methods that would help make its training better and more efficient.
The two worked together to bring a training method called Population Based Training (PBT for short) to bear on Waymo’s challenge of building better virtual drivers, and the results were impressive — DeepMind says in a blog post that using PBT decreased by 24% false positives in a network that identifies and places boxes around pedestrians, bicyclists and motorcyclists spotted by a Waymo vehicle’s many sensors. Not only that, but is also resulted in savings in terms of both training time and resources, using about 50% of both compared to standard methods that Waymo was using previously.
To step back a little, let’s look at what PBT even is. Basically, it’s a method of training that takes its cues from how Darwinian evolution works. Neural nets essentially work by trying something and then measuring those results against some kind of standard to see if their attempt is more “right” or more “wrong” based on the desired outcome. In the training methods that Waymo was using, they’d have multiple neural nets working independently on the same task, all with varied degrees of what’s known as a “learning rate,” or the degree to which they can deviate in their approach each time they attempt a task (like identifying objects in an image, for instance). A higher learning rate means much more variety in terms of the quality of the outcome, but that swings both ways — a lower learning rate means much steadier progress, but a low likelihood of getting big positive jumps in performance.
But all that comparative training requires a huge amount of resources, and sorting the good from the bad in terms of which are working out relies on either the gut feeling of individual engineers, or massive-scale search with a manual component involved where engineers “weed out” the worst performing neural nets to free up processing capabilities for better ones.
What DeepMind and Waymo did with this experiment was essentially automate that weeding, automatically killing the “bad” training and replacing them with better-performing spin-offs of the best-in-class networks running the task. That’s where evolution comes in, since it’s kind of a process of artificial natural selection. Yes, that does make sense — read it again.
In order to avoid potential pitfalls with this method, DeepMind tweaked some aspects after early research, including evaluating models on fast, 15-minute intervals, building out strong validation criteria and example sets to ensure that tests really were building better-performing neural nets for the real world, and not just good pattern-recognition engines for the specific data they’d been fed.
Finally, the companies also developed a sort of “island population” approach by building sub-populations of neural nets that only competed with one another in limited groups, similar to how animal populations cut off from larger groups (i.e. limited to islands) develop far different and sometimes better-adapted characteristics versus their large land-mass cousins.
Overall, it’s a super interesting look at how deep learning and artificial intelligence can have a real impact on technology that already is, in some cases, and will soon be even much more, involved in our daily lives.
Hardware developers toying with the idea of building physical stuff that can plug into the decentralized world of blockchain should point their eyes at Elk: A dev board in the making that’s been designed to support all sorts of IoT projects with a blockchain flavor.
Such as, for example, a connected door-lock that doesn’t demand that your ability to access your own property be dependent on the uptime (and accord) of servers of a remote corporate giant, nor your comings and goings be logged by a commercial third party.
Or, in another of their suggested examples, an alarm clock that charges you in bitcoin if you hit the snooze button too much, rather than getting up. Ouch.
The team behind Elk have just launched a crowdfunding campaign on Kickstarter to bring their prototype to market — with the aim of shipping the board to developers from next Spring.
They’re looking to raise a modest ~$20k. While the gizmo is being priced at $59 for early bird backers, or ten dollars extra for those who failed to, uh, un-snooze their clocks in time.
We covered Elk last year — when it was in an earlier stage of development and being called Elkrem.
At that point the team hoped to get the device to market before the end of the year. As it turns out it’s taken them a little longer to feel ready to fire up a crowdfunder — hitting various challenges along the way.
It’s worth flagging it’s not the team’s first product for hardware devs. They grabbed attention at TechCrunch Disrupt Europe back in 2013, when they got plucked out of startup alley as an audience choice to participate in our startup battlefield competition — where they pitched their idea to tap into sensors on smartphones as an alternative to Ardunio shields.
They went on to crowdfund and ship the 1Sheeld — and are still selling it to this day.
So there are fewer caveats than can usually apply to a crowdfunded hardware (though, as ever with anything being pitched for sale when still a prototype, it’s always prudent to expect delays).
Here’s a quick Q&A with Elk CEO and co-founder Amr Saleh on the team’s aim and ambition for the device:
TC: What is Elk and what is it for?
Saleh: Elk is a hardware development board for the blockchain and the decentralized web. It combines the simplicity of Arduino with native support for decentralized networks. With only a few lines of code you can build IoT that interfaces with Ethereum, IPFS, Whisper, and more!
Elk empowers developers to build what we call “Decent IoT”. Decent IoT is decentralized, gives users true control and true privacy, and allows entirely new use-cases like payments, oracles, selling your data, and much more.
With Elk you can build a smart door lock that you can control remotely without relying on a cloud provider that tracks and controls your device usage, or build a charging station that you can rent with Ethereum, or lock money into a treadmill that you can only get back when you work out. The possibilities are truly endless.
TC: Why is dedicated hardware necessary for developing blockchain IoT devices? What advantages does the hardware offer over rival dev boards, for eg, using microprocessors like Raspberry Pi?
Saleh: You can certainly use a microprocessor like Raspberry Pi to develop blockchain IoT devices. What differentiates Elk hardware-wise is that we combine both a microcontroller a microprocessor, a WiFi module and persistent storage preloaded with our OS in one breadboard-compatible board, and this allows us to offer a development experience that is plug-and-play just like programming an Arduino.
Unlike using a Raspberry Pi, with Elk you won’t have to deal with wallet and keys management, fuss over setting up nodes, tune their parameters to run well on an embedded device, handle crashes, etc. We are delivering the 10x easier Arduino-like experience to blockchain IoT development, with all the libraries that Arduino already supports. Developers can now focus on their applications and not the overheads.
TC: Who is the Elk for? How large is the blockchain hardware development community right now & how do you see that evolving over the next few years?
Saleh: Currently, blockchain hardware development is small and mostly siloed to building hardware wallets for blockchain enthusiasts.
We believe the potential for blockchain and decentralization extends far beyond that. Elk is not just for blockchain enthusiasts, but for privacy-conscious makers as well. Decentralization allows us to build IoT that is far more private, far more secure, and far more capable. We call it “Decent IoT”, and that’s what we are set out to introduce with Elk.
Current IoT architecture relies on centralized cloud providers for communication and data storage. This, by necessity, means that cloud providers (and whomever hacks them) can control your devices, deny you access, or tap into your private life.
The new decentralized web enables a completely new paradigm for IoT. A paradigm where your communication flows privately through a decentralized network with no central authority responsible for relaying your communication, no third party that can track your device usage, and no third party that can control your device. It additionally opens the door for other possibilities like payments, oracles, selling your data, and more.
Elk provides the tools and the UX to make building Decent IoT as easy as writing a few lines of code, and we’re hoping that over time this would further drive adoption of decentralization within the hardware community.
TC: Why the delay in launching the KS? What challenges have you encountered as you’ve prototyped Elk & how confident are you of meeting your estimated shipping deadlines?
Saleh: Blockchain and decentralization are very nascent fields, and ensuring that Elk offers the stable plug-and-play experience we want to offer was certainly a challenge.
Another significant challenge we faced was finding the right balance of features to offer in Elk. For example, we initially felt it was paramount for Elk to have a secure hardware enclave and spent months building out a prototype. We decided to later drop hardware security in favor of a stable and superior development experience. The development experience in building Decent IoT, we think, is far more of a bottleneck than pushing the extra mile in security.
At this point, we’ve been through four different iterations of our hardware and have done our diligence to be confident that we can deliver the product we’re offering with no surprises in production. We’ve already been through the process of manufacturing hardware. In our previous Kickstarter we shipped on time to our backers and sold tens of thousands of units in the years that followed.
TC: What’s the business model? Are you intending to make money via distributing/supporting the SDK as well as selling dev hardware?
Saleh: At this point, we are focused on making Elk the standard for building blockchain IoT devices. Beyond the current campaign, we’d be looking at enterprise use-cases that require stricter hardware requirements and support.
Autonomous delivery company Udelv has signed yet another partner to launch a new pilot of its self-driving goods delivery service: Texas-based supermarket chain H-E-B Group. The pilot will provide service to customers in Olmos Park, just outside of downtown San Antonio where the grocery retailer is based.
California-based Udelv will provide H-E-B with one of its Newton second-generation autonomous delivery vehicles, which are already in service in trials in the Bay Area, Arizona and Houston providing deliveries on behalf of some of Udelv’s other clients, which include Walmart among others.
Udelv CEO and founder Daniel Laury explained in an interview that they’re very excited to be partnering with H-E-B, because of the company’s reach in Texas, where it’s the largest grocery chain with approximately 400 stores. This initial phase only covers one car and one store, and during this part of the pilot the vehicle will have a safety driver on board. But the plan includes the option to expand the partnership to cover more vehicles and eventually achieve full driverless operation.
“They’re really at the forefront of technology, in the areas where they need to be,” Laury said. “It’s a very impressive company.”
For its part, H-E-B Group has been in discussion with a number of potential partners for autonomous deliver trials, and according to Paul Tepfenhart, SVP of Omnichannel and Emerging Technologies at H-E-B, but it liked Udelv specifically because of their safety record, and because they didn’t just come in with a set plan and a fully formed off-the-shelf offering – they truly partnered with HEB on what the final deployment of the pilot would look like.
Both Tepfenhart and Laury emphasized the importance of customer experience in providing autonomous solutions, and Laury noted that he thinks Udelv’s unique advantage in the increasingly competitive autonomous curbside delivery business is its attention to the robotics of the actual delivery and storage components of its custom vehicle.
“The reason I think we’re we’ve been so successful, is because we focused a lot on the delivery robotics,” Laury explained. “If you think about it, there’s no autonomous delivery business that works if you don’t have the robotics aspect of it figured out also. You can have an autonomous vehicle, but if you don’t have an automated cargo space where merchants can load [their goods] and consumers can unload the vehicle by themselves, you have no business.”
Udelv also thinks that it has an advantage when it comes to its business model, which aims to generate revenue now, in exchange for providing actual value to paying customers, rather than counting on being supported entirely through funding from a wealthy investor or deep-pocketed corporate partners. Laury likens it to Tesla’s approach, where it actually has over 500,000 vehicles on the road helping it build its autonomous technology – but all of those are operated by paying customers who get all the benefits of owing their cars today.
“We want to be the Tesla of autonomous delivery,” Laury said. “If you think about it, Tesla has got 500,000 vehicles on the road […] if you think about this, for of all the the cars in the world that have some level of automated driver assistance (ADAS) or autonomy, I think Tesla’s 90% of them – and they get the customers to pay a ridiculous amount of money for that. Everybody else in the business is getting funding from something else. Waymo is getting funding from search; Cruise is getting funding from GM and SoftBank and others, Nuro is getting funding from SoftBank. So, pretty much everybody else is getting funding from a source that’s a different source from the actual business they’re supposed to be in.”
Laury says that Udelv’s unique strength is in the ability the company has to provide value to partners like HEB today, through its focus on robotics and solving problems like engineering the robotics of the loading and customer pick-up experience, which puts it in a unique place where it can fund its own research through revenue-generating services that can be offered in-market now, rather than ten years from now.
TC Sessions: Mobility on July 10 in San Jose is fast approaching. Get ready for a superb lineup of speakers like Dmitri Dolgov (Waymo), Eric Allison (Uber) and Summer Craze Fowler (Argo AI). See the full agenda here.
In addition to the outstanding main stage content, TechCrunch is proud to partner with today’s leading mobility players for a full day of breakout sessions. These breakout sessions will give attendees deeper insights into overcoming some of mobility’s biggest challenges and answering questions directly from today’s industry leaders.
Breakout Session Lineup
How much data is needed to make Autonomous Driving a Reality?
Presented by: Scale AI
We are in the early days of autonomous vehicles, and what’s necessary to go into production is still very much undecided. Simply to prove that these vehicles are safer than driving with humans will require more than 1 billion miles driven. Data is a key ingredient for any AI problem, and autonomy is the mother of all AI problems. How much data is really needed to make autonomy safe, reliable, and widespread, and how will our understanding of data change as that becomes a closer reality? Sponsored by Scale AI.
Think Big by Starting Small: Micromobility Implications to the Future of Mobility
Presented by: Deloitte
A host of new micromobility services have emerged to address a broader range of transportation needs – bikesharing, electric scooters and beyond. The urban emergence of micromobility offers powerful lessons on finding the right balance between fostering innovations that will ultimately benefit consumers and broader transportation systems, while safeguarding public interests. Sponsored by Deloitte.
If You Build It, Will They Buy? – The Role of the FleetTech Partner in the Future Mobility Ecosystem with Brendan P. Keegan
Presented by: Merchants Fleet
The future will bring a convergence of new technologies, services, and connectivity to the mobility space – but who will manage and connect it all? Explore how FleetTech is creating the mobility ecosystem to help organizations embrace technologies – adopting your innovations through trials and pilots and bringing them to market. Sponsored by Merchants Fleet.
The Economics of Going Electric: Constructing NextGen EV Business Models
Presented by: ABB
How do we make the rapidly growing EV industry operational and scalable? Join ABB, HPE and Microsoft for a discussion on how government, industry, providers and suppliers are addressing market shifts and identifying solutions to build successful business models that support the future of mobility. Moderated and sponsored by ABB.
Bringing Efficiency to Closed-Course AV Testing with Atul Acharya
Presented by: AAA Northern California, Nevada & Utah
Looking to jump-start or accelerate your automated vehicle test operations? AAA has built its expertise by operating GoMentum Stations and performing safety assessments on multiple AVs and proving grounds. Join AAA as it shares its collective technical and operational learnings and testing results that will bring efficiency to your testing efforts. Sponsored by AAA Northern California, Nevada & Utah.
Friction-Free Urban Mobility
Presented by: Arrive
What does the future of seamless, urban mobility look like? How do mobility-as-a-service providers and connected vehicles work together to power transportation in a smart city? And which platform will aggregate all of the providers? In what promises to be a thought-provoking discussion, Arrive’s COO Dan Roarty will lay the foundation for what a city’s connected future will look like and outline key steps needed to achieve it. Sponsored by Arrive.
Michigan’s Mobility Ecosystem
Presented by: PlanetM
Revolutionary things can happen when some of the brightest minds in technology come together in one room. This Breakout Session will offer key insights into Michigan’s mobility ecosystem: the people, places and resources dedicated to the evolution of transportation mobility. Following a brief discussion, attendees will have the opportunity to connect with the people and companies moving the world forward through technology innovation and collaboration. Sponsored by PlanetM.
Autonomous driving company Waymo has launched its tie-in with Lyft, using a “handful” of vehicles to pick up riders in its Phoenix testing zone, per CNBC. To be eligible, Lyft users requesting a ride have to be doing a trip that both starts and ends in the area of Phoenix that it’s already blocked for for its own autonomous testing.
The number of cars on the road is less than 10, since Waymo plans to eventually expand to 10 total for this trial but isn’t there yet. Those factors combined mean that the number of people who’ll get this option probably isn’t astronomical, but when they are opted in, they’ll get a chance to decide whether to go with the autonomous option via one of Waymo’s vans (with a safety driver on board) or just stick with a traditional Lyft .
Waymo and Lyft announced their partnership back in May, and the company still plans to continue operating its own Waymo One commercial autonomous ride-hailing service alongside the Lyft team-up.
MIT’s Computer Science and Artificial Intelligence Lab has developed a new deep learning-based AI prediction model that can anticipate the development of breast cancer up to five years in advance. Researchers working on the product also recognized that other similar projects have often had inherent bias because they were based overwhelmingly on white patient populations, and specifically designed their own model so that it is informed by “more equitable” data that ensures it’s “equally accurate for white and black women.”
That’s key, MIT notes in a blog post, because black women are more than 42 percent more likely than white women to die from breast cancer, and one contributing factor could be that they aren’t as well-served by current early detection techniques. MIT says that its work in developing this technique was aimed specifically at making the assessment of health risks of this nature more accurate for minorities, who are often not well represented in development of deep learning models. The issue of algorithmic bias is a focus of a lot of industry research and even newer products forthcoming from technology companies working on deploying AI in the field.
This MIT tool, which is trained on mammograms and patient outcomes (eventual development of cancer being the key one) from over 60,000 patients (with over 90,000 mammograms total) from the Massachusetts General Hospital, starts from the data and uses deep learning to identify patters that would not be apparent or even observable by human clinicians. Because it’s not based on existing assumptions or received knowledge about risk factors, which are at best a suggestive framework, the results have so far shown to be far more accurate, especially at predictive, pre-diagnosis discovery.
Overall, the project is intended to help healthcare professionals put together the right screening program for individuals in their care and eliminate the heartbreaking and all-too common outcome of late diagnosis. MIT hopes the technique can also be used to improve detection of other diseases that have similar problems with existing risk models with far too many gaps and lower degrees of accuracy.
Argo AI will invest $15 million over five years to create a center for autonomous vehicle research at Carnegie Mellon University, one of the latest efforts by the Ford-backed company to accelerate the development of self-driving cars.
The center, Carnegie Mellon University Argo AI Center for Autonomous Vehicle Research, will focus on advanced perception and decision-making algorithms for autonomous vehicles, the company said Monday.
The investment follows the introduction of Argoverse, a set of curated data and high-definition maps that Argo AI released for free to researchers. Argoverse was created to give academic researchers the ability to study the impact that HD maps have on perception and forecasting, such as identifying and tracking objects on the road, and predicting where those objects will move seconds into the future.
Argo sees Argoverse and now this research lab as ways to encourage more research and hopefully breakthroughs in autonomous vehicle technology.
“We are thrilled to deepen our partnership with Argo AI to shape the future of self-driving technologies,” CMU President Farnam Jahanian said in a statement. “This investment allows our researchers to continue to lead at the nexus of technology and society, and to solve society’s most pressing problems. Together, Argo AI and CMU will accelerate critical research in autonomous vehicles while building on the momentum of CMU’s culture of innovation.”
Argo’s investment in CMU makes sense. Argo’s headquarters aren’t far from CMU. And the university is known for its robotics program. There’s also a personal connection.
Argo was founded by a team with deep CMU roots. Co-founder and president Peter Rander earned his masters and PhD degrees at CMU. Rander and Argo AI CEO Bryan Salesky worked together for years at National Robotics Engineering Center, a unit within Carnegie Mellon University’s Robotics Institute.
Rander left and became the engineering lead at Uber ATG and Salesky went over to the Google self-driving project, now called Waymo. They left their respective jobs to form Argo in 2016.
This isn’t the first autonomous vehicle company to see potential in CMU.
In 2015, Uber announced a strategic partnership with CMU that included the creation of a research lab near campus aimed at kick starting autonomous vehicle development. But that relationship ended up gutting CMU’s own robotics lab known as as the National Robotics Engineering Center. Before the year was up, dozens of people, including the NREC’s director, had left to work at the Uber Advanced Technologies Center.