Four years ago, Comma.ai founder George Hotz turned to his board — of which he is the sole member — and fired himself as CEO. At the time, the goal of the famous iPhone and PlayStation 3 hacker, known as geohot, was to create a new research division to focus on behavioral models capable of driving cars.
Now, Hotz says he’s stepping away “from the time” of the driver assistance system startup that promises to bring Tesla Autopilot-like functionality to your car. However, he will remain its only board member and president.
Hotz hasn’t been involved in most day-to-day leadership duties for some time, he told TechCrunch. That goes to COO Alex Matzner and CTO Harald Schäfer. The company hasn’t had a CEO since 2019, when Riccardo Biasini held the position. (Biasini left the CEO position in 2019 and remained at Comma to work on its open pilot software until February 2020.)
Hotz was what Matzner described as an occasional observer and solver of difficult problems.
Comma.ai, which developed and now sells a $1,999 driver assistance system development kit compatible with more than 200 vehicles, is going nowhere, Hotz told TechCrunch. The focus now is on transforming the devkit, which runs on Comma’s open-source software called openpilot, into a productized consumer product.
“I’m good at wartime stuff,” Hotz told TechCrunch in a recent interview. “I’m not that good at practice, ok, let’s patiently expand that. “Do you want to do business with a supply chain capable of manufacturing 100,000 devices a year?” Like, not really.
And that’s one of the goals: annual sales of 100,000 Comma 3 units.
The startup quietly raised $10 million from individuals last year and moved into a 20,000 square foot facility in San Diego. (Comma’s initial funding of $8.1 million was raised in two rounds from Silicon Valley VC a16z.) It is now “aggressively recruiting” and is on track to launch upgrades. major end-to-end machine learning updates on openpilot later this month, Matzner told TechCrunch. in a recent email.
Comma.ai initially launched a plan to sell a $999 self-driving car kit that would give certain vehicle models highway driving assistance capabilities similar to Tesla’s Autopilot feature. Hotz canceled those plans in October 2016 after receiving a letter from the National Highway and Traffic Safety Administration. Five weeks later, Comma.ai released its self-driving software to the world. All code, along with hardware blueprints, has been published on GitHub.
The company has continued to develop an ecosystem of hardware products all aimed at bringing semi-autonomous driving capabilities to cars. These efforts resulted in the Comma 3, which is priced between $1,999 and $2,499 depending on storage size. The car harness, which connects the dev kit to the vehicle, costs an additional $200.
Comma 3 is much easier to use than its previous iterations. Installation and setup takes a bit of patience, but no longer requires any technical expertise, Hotz said. Now it’s up to the company to take the Comma 3 and make it a “produced” and scalable consumer product, he added.
And after?
Hotz is already deep into his next project, which he calls Tiny Corporation. Its goal is to write a new framework for machine learning that is faster and less complex than PyTorch. Instead of training the ML model in the cloud and shipping it to the edge, Hotz wants to build tools to train ML models at the edge.
“The current PyTorch and TensorFlow aren’t going to be enough to form the edge,” he said.
AI-related fields, including automated driving, are leaning more towards deep neural networks – a sophisticated form of artificial intelligence algorithms that allow a computer to learn by using a series of connected networks to identify patterns in the data… a kind of brain function. But as Hotz notes, “we’re all pretty new to this neural network stuff.”
Andrej Karpathy, a deep learning and computer vision expert and former AI director at Tesla, called this programming step 2.0, or Software 2.0, in which programming is done by example and humans are really just writing the general scaffolding. In other words, software that writes itself.
“You shouldn’t build an (AI) chip until you can build software that outperforms or at least performs the same as PyTorch on Nvidia,” Hotz said. “Waiting for AI chips to be built, let’s build the software first.”