Published on December 19th, 2015 | by Kyle Field58
Geohot Takes On Tesla Autopilot From His Garage
December 19th, 2015 by Kyle Field
When the iPhone first came out, it was carrier locked to AT&T, meaning that anyone locked into a contract with T-Mobile or another carrier couldn’t use the iPhone. It’s a far cry from the situation today, where the phone is available on pretty much every carrier on almost every continent, but back then, it was frustrating. George Hotz, a 17-year-old kid from Jersey, was frustrated enough that he decided to dedicate his summer to learning how to reverse engineer the phone in order to carrier unlock it for use on T-Mobile… and after 2 months of nonstop effort, a scraggly George Hotz was successful in August of 2007, going into the record books as the first human to carrier unlock an iPhone.
I myself utilized several of Geohot’s (his online handle) software-based carrier unlocking tools later in the development cycle and really enjoyed his desire and ability to quickly assimilate and utilize extremely technical tools to solve unique, non-traditional problems. Geohot similarly challenged authority when he was the first to hack the encryption on the Sony Playstation 3, for which he was later sued. His seemingly ceaseless desire to challenge authority was put on stage front and center, and after crowdfunding his defense money, George settled out of court with Sony and moved on to his next adventure.
Mr Hotz’s LinkedIn profile reveals a turbulent history working short stints at many of the big names in Silicon Valley, with companies like Google, Facebook, SpaceX, and Vicarious that tenuously hang on to the bullet points of his resume as if they knew it was temporary when he first walked through the door. Specifically, geohot worked on the Google Street View program, where he worked to build the next generation of the hardware–based image capture system used by the software to gather images, integrating feeds from up to 25 cameras.
Flashing forward a few years, he was back to his independent roots as a freelance Chief Reverse Engineer, under which he notes that he was a self-made millionaire before turning 21. Throughout his resume history, it is clear that hardware hacking is his obvious passion, starting with the iPhone, then to Google, and even working at SpaceX, where he designed and built an HDMI adapter for an LCD screen. I read the frequent turnovers of employment as his resistance to conform to even the most liberal of corporate cultures (Google, Facebook). The top-tier companies he has worked for make it clear that George is extremely intelligent and a quick study of new tech… but that is counter balanced with a short attention span.
The driving force in George’s is his intense, unquenchable desire to learn, which fuels his ongoing search for a challenge worthy of his skills. Layer on top of that, cutting-edge hardware and imaging project work at industry-leading innovator Google. Never satisfied, he stepped up to what should have been a tougher challenge at SpaceX, soaking in the fast-paced startup mentality of the well-funded private rocketry company. Looking towards the future, George explored on his own and after quite a bit of solo research, joined the ranks of the cutting-edge AI startup Vicarious, where he helped develop next-generation AI algorithms as a researcher. We are now seeing the first fruits of this next phase in the evolution that is Geohot, and it is evident that this latest venture has him excited. He hasn’t previously ventured so far as to start up a serious venture like this, which has prospects for real profit. So, we may see yet another evolution of the tech mastermind that is Geohot as he tries on a new hat, as a businessman.
A few weeks back, George reached out to Ashlee Vance, the shoot-from-the-hip author who recently penned Elon Musk’s biography (and now writes for Bloomberg). George was eager to show off everything that he had been working on from his garage since leaving Vicarious. The fruits of his labors sat next to his desk in the garage — a white 2016 Acura ILX which looks like a modern-day Frankenstein with wires popping out of every orifice and a rotating LIDAR array up top. Hotz shared that he cobbled the creation together in right around 1 month from start to finish, only starting the build in late October and first showing it off to Vance in late November. That’s insanely quick after seeing how much he has done with it — technically speaking.
Hotz pulled together all the tricks learned from his previous employers, rolled together with his own formal research into AI during a brief stint at Carnegie Melon, where he aced a few AI classes and learned all the institution had to teach him on the subject. Recalling his time at the college, where he voraciously consumed tremendous amounts of information on AI, both in and out of the classroom, Hotz relates to Vance, “The math is simple. For the first time in my life, I’m like, ‘I know everything there is to know.’ ” There’s obviously no shortage of confidence in this kid, but given all that he’s already done and all of the previously “impossible” tasks he has already conquered, why not throw some faith behind him?
So, what can Hotz’s car do today, and how does it compare to a Model S? The real head-to-head test is yet to be seen, but Hotz took Ashlee out for a spin in the AI-enabled ILX and Vance was clearly impressed. Over 3 different visits over a period of weeks ending in early December, Vance was able to see the accuracy of the path that the car thought it should go and the path it was traveling come to be nearly in sync, revealing that the AI learning component of the system has a much better idea as to where it should be from watching others — both the driver of the car and other objects on the road.
What’s unique about a full AI-based solution is that it does not look for rules to dictate where the car can or cannot drive, but rather, it learns as a human learns — by watching and remembering… the difference is that a machine learns a lot more. It was this same type of AI-learning algorithm that allowed Tesla’s Autopilot to improve — but at a much faster pace. Tesla has the unique advantage in that it has a few thousand cars out on the roads that it can pull data from, sharing the data back out to the masses. Every day, the combined system gets smarter. It learns as one car learns. Each correction, each turn signal, and each manual intervention is a lesson learned… for all of the fleet.
Vance notes in the article that Hotz is also very proud of how efficient his code is — utilizing right around 2000 lines of code. He attributes this efficiency not to the fact that he knows everything (I joke!) but to his depth of understanding of the fundamentals of AI and uncanny insight into how it learns: “We’ve figured out how to phrase the driving problem in ways compatible with deep learning.”
So what? What does Hotz’s plan to do with his new Frankenacura? Take over the world, obviously. Hotz does have big aspirations and plans to develop several business models from his most recent creation, sharing that he can see himself running a dealership selling autos fitted with his tech package or even selling the package to trained installers who can embed the system in any compatible car. The Bloomberg article elaborates on the systems being used:
“Hotz plans to best the Mobileye technology with off-the-shelf electronics. He’s building a kit consisting of six cameras—similar to the $13 ones found in smartphones—that would be placed around the car. Two would go inside near the rearview mirror, one in the back, two on the sides to cover blind spots, and a fisheye camera up top. He then trains the control software for the cameras using what’s known as a neural net—a type of self-teaching artificial-intelligence mechanism that grabs data from drivers and learns from their choices. The goal is to sell the camera and software package for $1,000 a pop either to automakers or, if need be, directly to consumers who would buy customized vehicles at a showroom run by Hotz. ‘I have 10 friends who already want to buy one,’ he says.”
Hotz is taking a new angle on a market that is heating up as standalone vehicle automated safety companies like MobileEye begin competing for business with and giant, established auto manufacturers like Mercedes-Benz, Nissan, Volvo, Tesla, and many others who are charging into the space. Who will dominate will be decided by the market, but for now, Tesla is leading the pack in terms of implemented products.
With bold plans to show Tesla up in a faceoff in NorCal on the infamous Golden Gate Bridge as well as in the famed, ever-gridlocked 405 freeway in Southern California, Geohot’s new adventure is indeed heating up. I’m excited to have another player in the game, as competition will only drive these uber-competitive CEOs on to bigger and better things… I hope.
Check out the official Bloomberg coverage of time spent with George and his car:
Have a tip for CleanTechnica, want to advertise, or want to suggest a guest for our CleanTech Talk podcast? Contact us here.