DIY autonomous driving on public roads must stop NOW

Some guy just shared how he hacked his android phone to drive his car (Web Archive here).

Cool. But if I get that post right, that project is brilliant, necessary AND ran in a really stupid, irresponsible way. Therefore, I really hope social pressure, before anything else, will “reset” (not cancel!) every project like that as soon as possible. Here's why.

"Responsibly with proper safety and caution"

Advanced driver-assistance systems (ADAS) are "electronic technologies that assist drivers in driving and parking functions".

The goal of that project is "to make a [smartphone-based] ADAS system that keeps you in lane and adjusts speed of the car based on lead vehicles".

The vision is (comment from a co-developer/alfa tester) "Imagine going to a local shop, buying some cables, downloading an android app and getting access to a world class autopilot system".

This is the beginning of the post:

WARNING: All the stuff shown here has been done responsibly with proper safety and caution. Follow at your own risk. This project is NOT for you if you do not understand what’s happening.

This blog post is an overview journey of me and friends making an affordable self driving car system with 2 constraints:

• All the stuff that we use to build it should be available in local shops near us.

• The system should drive 90% of the time without manual intervention between two cities.

The good part

I am all for Citizen Science, freedom to hack and Right to Repair. I also have nothing against adults risking their own lives to do whatever is really important for them, just "because it's there".

Developing and testing do-it-yourself autopilot systems can be great fun, a great way to learn programming and other technologies. Besides, it can explore really innovative solutions that traditional carmakers may never try, or even imagine. Not to mention that, to the extent that autonomous driving is really needed and usable, converting existing, still good cars should be much, much greener and cheaper than making new ones. So yes, if you find it interesting and can do it, please do it.

The damn stupid part

Hacking smartphones to drive cars is a great idea... as long as the driving only happens on closed tracks.

Driving that way on public roads, instead, is terribly stupid and irresponsible.

The “NOT for you if you do not understand” part of that mission statement is valid only in a universe where the Dunning-Kruger effect does not exist, but never mind that.

The real problem is the "Follow at YOUR own risk" part, which is so false and disconnected from reality that I can't figure out if it was written in good or bad faith, and I don't know which would be worse.

Source: Reporter reveals ‘corners cut’ on Titan

DIY autopilots made with "stuff that should be available in local shops near us" are as safe and smart as uncertified submersibles driven with cheap game controllers (yes, it's TRUE). Just much worst because, unlike the submersible, they could kill other people. Or cripple them, in a way that no insurance company will ever refund. As it is now, that project is "Follow at OTHER, TOTALLY INNOCENT PEOPLE's risk".

Uncertified, DIY autopilots are great. The more, the better. Just NEVER on public roads.

But they know what they are doing!

Even if that were true, so what? Almost every driver who hit a pedestrian surely thought the same thing until a moment before the crash. Including Rafael Vasquez and his managers:

Above all, these are projects that everybody can download and use in the same way as its developers.

Surely, the current developers and first adopters are not average people. So, for all I know they may, indeed, be all so good to never fall asleep or get distracted as the rest of us, ever. Theoretically.

But the second wave of adopters, and every wave after that, will surely include idiot or seriously incompetent drivers.

Like, you know, those apexes of the human species who one week ago killed a 5-year old child in Italy, because they thought that spending 50 hours mostly driving a car so powerful they couldn't even legally rent it was a cool idea:

Without that accident, sooner or later those geniuses would have very likely tried "spending 50 hours in a car driven by my smartphone".

Experimental vehicles, especially if developed by amateurs, belong to test tracks and nowhere else.

Uncertified, DIY autopilot systems on public roads are a really stupid idea that cannot scale and must stop as soon as possible. Please share this request as much as you can and tell mansin, flowdrive_ai and every openpilot developer to continue, but NOT on public roads, ever.