Geohot’s new automated-driving device can only be redeemed by coughing up data

AUSTIN, Texas—At the first day of the SXSW Interactive festival, George “Geohot” Hotz announced an updated business plan for his company Comma.ai, all while revealing a new piece of hardware that will be given away, as opposed to sold: the Panda.

The small circuit-board device comes with an ODB2 connector on one end and a USB port on the other. Hotz described the ODB2 as compatible with any car made after 1996, though ideal for cars made later than 2006. The device’s crowded circuit board also includes a 32-bit processor, a Wi-Fi driver, and a 4A charger, which he described as “an awesome phone charger.” Hotz said the Panda can expose more active car data than the ODB2 plug-in devices used by apps such as Torque, including individual wheel speeds, steering wheel angles, and blinker functions.

The ODB2 even has the ability to issue accelerator and brake commands to a car.

How do you get a Panda? Simple: Give Hotz all of your data. Comma.ai launched a dash-cam app late last year, called Chffr (short for “chauffeur”). Today, Comma.ai announced that the app’s 20 most avid users every week will earn “Comma points.” Earn enough of those, Hotz said, and you can trade them in for your very own Panda. (Users cannot currently spend money to get a Panda, Hotz told Ars, and he couldn’t answer whether that may change in the future.)

Hotz said that his company’s new business plan revolves around aggregating driving data from users across the world—in particular, Comma will combine GPS and camera footage to accurately map and track a huge swath of driving patterns. The company will eventually make that data available as part of a service that will be called “Comma Premium.”

“You’re all cool, so I’ll let you in on my real plan,” Hotz told the crowd. “I’ll run [a free, autonomous driving app] with ads so annoying that you’ll subscribe to Comma Premium for $24 per month. Here’s a picture of a yacht. Now I’ll take questions.” This blunt statement followed his suggestion that future autonomous driving systems (his or others) might go so far as to automatically drive people to advertisers’ businesses on their way elsewhere. His example was a forced pit stop at Taco Bell on the way to a grandmother’s house.

Comma One “didn’t make sense at that scale”

Hotz said his company would focus on “replacing the software” that runs car driving systems, as opposed to becoming a hardware company. This, he emphasized, was the logical course of action after receiving pushback from the NHTSA in October 2016 over his company’s planned Comma One device.

That device would have been plugged into compatible cars to enable “level 2” driving autonomy (level 2 devices manage the acceleration, braking, and steering of a car with an aware passenger behind the steering wheel). But the NHTSA requested that the device receive testing trials before any commercial launch. Shortly after receiving the letter, Hotz canceled plans to launch the device.

“We could do that, go back and forth with NHTSA, and do the testing they were looking for, but it would’ve cost us a lot of money,” Hotz explained to the SXSW crowd. “The plan was only to sell 1,000 Comma Ones. [NHTSA testing] didn’t make sense at that scale. Our goal has always been to own the data and own the network. Whether we ship hardware here, ship open source software there… it doesn’t matter. After the letter came in, [Comma One] made less financial sense.”

Hotz explained to Ars Technica that he plans to eventually provide software to Panda owners that would enable level 2 driving autonomy. He did not clarify when or how that would roll out or whether that software would specifically be Chffr.

He reminded the crowd that enterprising users can become part of the Comma autonomous driving world right away, should they combine a OnePlus 3 Android smartphone with other off-the-shelf parts. Hotz directed attendees to visit Comma’s Github depository, where they can find the schematic necessary to build a Comma Neo device. It requires, among other things, roughly $700 of combined parts (including the OnePlus 3), an unlimited-data SIM card, a 3D printer, and quite a bit of soldering. Its Github home offers an all-caps, bolded warning that the project is “alpha-quality software for research purposes only.”

The same warning can be found at Comma’s OpenPilot open-source driving app, which currently “enables the functions of Adaptive Cruise Control (ACC) and Lane Keeping Assist System (LKAS) for Hondas and Acuras.” Hotz says more car compatibility is coming soon to OpenPilot, including a port for Toyota cars. He also says that the current fleet of “50 to 100” Comma Neo drivers will expand to “the second-largest network of self-driving cars” by the end of April—though that statement could mean a pretty small number, since he didn’t clarify types of roads or other qualifying information.

While Hotz’s talk was mostly about controlling and profiting from data that Comma.ai hopes to accumulate via its open-sourced software, he conceded that Comma’s own produced devices could become a possibility, especially as he perfected the software side of the autonomous driving equation. “Would you sell this thing to soccer moms?” he asked as he held up a home-built Comma Neo. “No way. But would you sell something like this in a shiny case at dealerships to soccer moms? Oh yeah.” He confirmed that he’d be more willing to work on tests with the NHTSA, “with lawyers and a plan,” at a scale of 10,000 to 100,000 devices.

When asked about his recent attempt to purchase a Tesla of his own, Hotz declined to specifically answer. “I like Tesla, we’re not going into that story,” he said. “Their fight’s not with me. They don’t realize it.”

Listing image by Sam Machkovech

Related articles

Comments

Share article

Latest articles