🚗 Who will win the race to build a self-driving car?
The maverick entrepreneur taking an open source approach.
In 2015, Elon Musk predicted full self-driving cars were just around the corner, speculating it would be just a couple of years before ‘complete self-autonomy’.
I don’t know about you, but I still drive myself around.
But not everyone does.
Cruise, a self driving car company owned by General Motors offers an entirely self driving taxi service in San Francisco as of June this year.
It hasn’t all been smooth sailing, with incidents such as when police officers pulled over a self driving car and were at a loss what to do. Also, the Cruise cars keep blocking traffic which is starting to annoy people.
In spite of these challenges, it is true that today you can call a self driving taxi in San Francisco and it will take you home, without any human intervention.
So is that it then?
Cruise has won the self driving car race?
Well, not quite.
Although Cruise has managed a remarkable achievement in setting up the first publicly available self-driving taxi service, we are still a way away from self-driving being a standard feature of the next car you buy.
Today I am going to take a look at the state of self-driving cars, and in particular at one underdog company taking a different approach.
2015 was a good time to be in self driving cars.
It was also a time when I was happened to be living near San Francisco, and I remember seeing the funny little Waymo cars driving around.
It seemed we were on the cusp of a self-driving revolution, with frequent predictions that self-driving cars were imminent.
This was a classic ‘Peak of Inflated Expectations’, as described by the Gartner Hype Cycle.
If 2015-2017 was the ‘Peak of Inflated Expectations’, today we are firmly in the ‘Trough of Disillusionment’, with articles coming out with titles like “Will self-driving cars on our roads ever be a reality? Some experts are becoming sceptical”.
In other areas, such as AI image generation, the pace of development has been quicker than even experts expected, but self-driving cars have been a harder nut to crack than many expected.
So which companies are dragging us out of the trough?
Well the first one I have already mentioned, Cruise. Cruise is planning to keep expanding its self-driving taxi service, but at the moment it doesn’t seem to be porting its software over to other cars. However, Cruise is owned by General Motors, so I am sure this is something they are thinking about.
Of course it’s impossible to talk about self-driving without talking about Tesla. Tesla already ships cars with ‘full self driving’ for a $5000 premium. Unfortunately, their Full Self Driving software is not really what a normal person would think when they hear ‘full self driving’, so Tesla is unsurprisingly being sued. And every year Elon predicts that full self driving is just around the corner, so again, this will be the year. We have learned by now not to hold Elon to his timelines.
Waymo, a spinout from Google, also offers its self driving taxi service in San Francisco, but just to a select group of beta testers largely made up of Google employees.
All these companies have their own, closed source approaches and either use many cameras (Tesla), or expensive LIDAR arrays (Waymo, Cruise).
Waymo, Cruise and Tesla are all big, multi-billion dollar companies, but there is one startup that is zigging when everyone else is zagging.
Comma.ai is a small company run by a maverick called George Hotz, better known online as “Geohot”.
Geohot is well known in tech circles as an extremely talented hacker - he was the first person to jailbreak an iPhone aged just 17.
As you might expect George is somewhat unorthodox, so Comma’s approach to self driving cars is a bit different to other companies.
In fact, he apparently turned down a job at Tesla with a “multi-million dollar bonus” in order to start Comma.ai.
Comma is an interesting company for many reasons, of course primarily for their approach to the self-driving car problem. But before we talk about that I am going to take a quick detour to share a couple of the quirky facts I discovered reading about Comma.
Ok so normally if you want to speak to sales, someone from that team is usually pretty happy to get on the phone with you. After all, that’s a great opportunity to sell you something. At Comma, the privilege of speaking with their sales team will cost you $1000 for 30 minutes. You can just go ahead and buy 30 minutes right on their website if you want.
Also, Comma recently got targeted by a patent troll company. Patent troll companies seek to buy up dormant patents and then extort companies through lawsuits. Quite often this works because firms would rather settle out of court rather than go through the expensive and time consuming trial process. It’s quicker and cheaper, so many companies make the calculation, and therefore the emboldened and richer patent troll company continues to shake down more and more real companies. George did not choose this route.
“What he didn’t take into account is that comma isn’t run by rational actors in suits sitting on a committee. It’s run by me, George Hotz. I’m willing to lose $1M before I give him $10k. We will hire an amazing legal team, fight this, and while doing so invalidate his patents so they can’t be used against anyone else. Not because it’s rational, but because it’s the right thing to do. No patent troll will ever get a dollar from comma.” - George Hotz
This is an approach I admire. You can read George’s whole response to the lawsuit here.
Anyway, digression over, back to cars.
So how is Comma’s approach to self driving different?
It starts right from their philosophy. Rather than try to build a new car, or equip a car with lots of expensive LIDAR arrays, they decided they would try to build self driving capabilities on top of the existing cars that people already own.
So they created a product that could be bought off the shelf and set up by basically anyone.
What does that mean exactly? Well for ~$3000 dollars you can buy the Comma 3, which is a smartphone sized device that you mount on your windscreen.
Interestingly the earlier models were actually smartphones.
Once setup, you can use Comma’s software called Openpilot, which uses existing factory-installed hardware to improve a vehicle's ability to center itself in a lane, change lanes and many other self driving features of increasing complexity.
The low cost of the device, plus the fact it is compatible with 200+ existing cars, means they are able to collect millions of miles of real world driving data at a much lower cost than many of their competitors.
For comparison, the cars that Cruise and Waymo use cost around $250,000.
The other major difference is that Comma’s software is open source, so if you are a software engineer and your car isn’t supported, you can fork the repository and make your own customisations.
This open source approach has spurred the development of a community of almost 30,000 people gathering in Comma’s Discord to discuss, hack and remix Comma.
Comma has tough competition. Tesla, Waymo and Cruise all have very deep pockets, and that’s before you even mention the legacy car companies.
But small, nimble teams have accomplished amazing things in many areas before. Whatsapp was sold for $19Bn when it had just 55 employees. There were 20 people on the AlphaGo paper. Amazing things can be done by smart and focussed people.
I don’t know who is going to ultimately win the race to build full self driving, but I do know who I hope will win.
Until next time,
Jamie
Did you enjoy this week's newsletter? Help me to improve!
Click on a link to vote:
Thanks for your feedback!