comma.ai, self-driving cars and indefinite optimism

comma.ai gets punished for being open, so why would anyone even bother?

I’ve blogged before about comma.ai and its founder George Hotz. So I was happy to see comma’s main product (basically a cell phone that can perform lane assist) on hacker news. But as is increasingly the case, the comment section was filled with a cynicism I wouldn’t normally expect from the tech community.

Many of the comments pointed out the, admittedly ridiculous terms of use:

THIS IS ALPHA QUALITY SOFTWARE FOR RESEARCH PURPOSES ONLY. THIS IS NOT A PRODUCT. YOU ARE RESPONSIBLE FOR COMPLYING WITH LOCAL LAWS AND REGULATIONS. NO WARRANTY EXPRESSED OR IMPLIED.

I view this as a knee jerk reaction by an immature founder. He’s had a similar response when someone once asked him about the privacy policy of all the video he collects. He snickered when asked and replied gleefully that by using comma 2, you give comma.ai the right to do anything they want with your video data.

To Hotz, he just doesn’t want to bother with anything like a guarantee or sensible thought out privacy policy. Every comma device records everything it sees and sends the raw video data (not just vectors) to comma.ai for learning. His goal is to solve self-driving while making shippable intermediaries. Your video data is critical to that goal.

I admit that there should be a sensible policy, but at least he’s honest. I wish Facebook just said “we do whatever we want with your data” instead of a 4,000+ word privacy policy.

How about the liability? Comma.ai abdicates any and all responsibility. But is this any different than the policy of other lane keeping technologies?

The Competition

Cadillac Super Cruise

Even while using the Super Cruise driver-assistance feature, always pay attention while driving and do not use a handheld device. Requires properly equipped vehicle, active Super Cruise subscription, working electrical system, cell reception and GPS signal.

NOTE: I couldn’t find an official statement of liability or terms of use for any of the competitors I researched. I was just able to find marketing copy. Maybe you have to buy the car and read the hard copy of the owners manual.

Another interesting note:

*Compatible Super Cruise roads for 2020 and 2021 model-year Super Cruise-enabled vehicles; dealer-installed software update required for 2018 model year and 2019 model year in order to be compatible with these roads.

No over the air updates, which is kind of weird since they charge you $25 a month service after your free trial is up. What version do you have anyway? 🤷

Also note:

The available Super Cruise driver-assistance feature is available only on compatible roads that are separated from opposing traffic.

So this product is the equivalent of one of the old school GPS units with maps that come on special CDs that have to be installed by a professional.

Ford Driver Assist

What… the… fuck…

I’m sorry, your model had “Reverse Brake Assist”, but not “Hill Start Assist”.

Who can look at something like this and think the software underlying this monstrosity is good? Isn’t it software? Why is it different for every car?

Audi Driver Assist

How does Audi Driver Assist work? Here’s a helpful artist rendition.

Want to view key info?

Cool.

This is the only information you can find on the other competitors. Images of scenic roads with a person driving with their hands on their lap. Bad artistic renderings of object recognition. A list of features® and the corresponding models the features® are available on.

I could go on, but let’s just say the industry is a mess.

comma.ai

What about comma.ai? Well, they have a real product and miles that blow the others out of the water. You know what software you’re running, or at least partially. You know how to get in touch with the developer team (I personally suggest yelling at Hotz on Twitch). You have an issue board and a lot of enthusiasts running the software. The competitors have none of this. They have a marketing brochure. That’s it.

But for some reason it’s cool to shit on people and companies that are working in important fields and doing unconventional things. Here’s a typing app that lets me type out public domain novels! Finally, the product that will end the malaise.

But here’s a product that’s relatively open and transparent about what it can do. It’s a company run by engineers that caters to other engineers (I’ve heard from Hotz before that something like 40% of comma users run custom software). Here’s a company that’s actually working to make a change in the world.

And what do the very people you’re trying to cater to say? They say it’s dangerous, despite it having more miles than any other company, apart from Tesla. They complain about waiving liability without being able to point to a liability policy of any competitor. They laugh because parts of it is written in python, despite having no clue about competitor’s software stack (certainly not python, lol!).

In short, they’re just being derisive. You see the same thing with people criticizing Tesla autopilot. Elon Musk is trying to save humanity and colonize Mars, but god forbid he disagrees with California covid-19 employer mandates! That’s the cardinal sin. Don’t you know? Only Gavin Newsom can save us.

Indefinite Optimism

Peter Thiel famously said:

We wanted flying cars, instead we got 140 characters

Part of the reason is that the smartest people aren’t working on the hardest problems. The siren song of inflated tech salaries is too much for some to resist. But it’s also this pervasive skepticism that has crept into our society.

Thiel writes about this in Zero to One.

Thiel saves his harshest criticism of indefinite optimism.

You can expect the future to take a definite form or you can treat it as hazily uncertain. If you treat the future as something definite, it makes sense to understand it in advance and to work to shape it. But if you expect an indefinite future ruled by randomness, you’ll give up on trying to master it.

Indefinite attitudes to the future explain what’s most dysfunctional in our world today. Process trumps substance: when people lack concrete plans to carry out, they use formal rules to assemble a portfolio of various options. This describes Americans today. In middle school, we’re encouraged to start hoarding “extracurricular activities.” In high school, ambitious students compete even harder to appear omnicompetent. By the time a student gets to college, he’s spent a decade curating a bewilderingly diverse résumé to prepare for a completely unknowable future. Come what may, he’s ready—for nothing in particular.

The indefinite optimist believes that self-driving technology will work. We just need an integrated approach coming from big auto, the result of billions in investment and R&D. They’ll figure it out.

The definite optimist believes the future will be better than the past if he works to make it better. This is essentially Hotz. Optimistic enough to devote his time to self-driving technology, but knowing that the breakthrough is not a certainty.

So let’s not mock those with the fortune and drive to work in meaningful fields that can shape the next generation. Let’s applaud those that open up their work, despite all its blemishes. Let’s elevate those that are brave enough to be optimistic about whats possible, but realistic enough to knows it’s not going to just happen.

But most importantly let’s just stop being so derisive when it comes to world changing technologies.