3 ms·
I've said this before, so apologies to those who've heard it already, but I believe testing big fast robots in public is irresponsible. These machines have alr
by carapace 5y ago
I've said this before, so apologies to those who've heard it already, but I believe testing big fast robots in public is irresponsible. These machines have already killed people. We know that human drivers kill and maim tens of thousands of people every year, so it's pretty obvious that robotic drivers will do so too, unless and until they become better drivers than humans. If somebody wanted to let their killer robots move around in public with limited supervision it would be obviously illegal, but somehow dressing up as cars falls into a weird blind spot.
I've encountered at least two major objections:
> Human drivers kill people so it's okay if robot drivers do too.
But is it really okay for human drivers to kill people? Our complacency in the face of automotive mayhem is the result of a deliberate campaign of social engineering: "The Real Reason Jaywalking Is A Crime" (Adam Ruins Everything) https://www.youtube.com/watch?v=vxopfjXkArM https://www.youtube.com/watch?v=vxopfjXkArM
We went from decrying "speed demons" (reckless drivers who hit pedestrians) to victim-blaming "jaywalkers". Now, for example, more Americans have been killed by cars than by all the wars we've fought. (And that's just counting collisions, not deaths from air pollution, etc.)
> The machines will eventually be safer than human drivers (a belief I share) but they need to practice in real world conditions to do it.
So build a model city, populate it with people willing to take the risks, and test them there. (This is setting aside the additional idea of doing extensive modeling in simulated cities.) It would be expensive but it would avoid killing people who haven't e.g. signed a release agreeing to be test dummies in the robot car experiments.
The obvious way to make large, heavy, fast robot cars is to start with small, light, slow robot cars. Imagine a self-driving golf cart covered in nerf (soft foam) that doesn't go faster than about 5mph. There are applications for that and it would be much more difficult for such a robot to injure or kill people.
Going straight for self-driving cars (and can we please call them "auto-autos"?) seems to me to be hubris.
- cto_of_antifa 5y agoThis is a great and well reasoned post and in general, thank you. I like the test city idea; at first I thought it was still a little risky but then I realized that this is probably one of those situations where you can't actually test or prove its efficacy without a "live-fire" environment. In medicine you famously can't conduct an experiment that knowingly exposes participants to a greater risk of death than the alternative of not participating, and i wonder if that's a useful ethical framework for this
- tim333 5y agoThough say for argument's sake you just want to reduce road deaths in total. Wikipedia has the annual world total at 1,350,000. Tesladeaths.com had confirmed Tesla autopilot deaths for 2021 at 10. On that basis you might want to crack ahead with developing self driving to bring the 1.3m figure down at some point.
- 6gvONxR4sf7o 5y agoThose numbers aren’t remotely comparable. Airline autopilot deaths are tiny compared to human pilots, but they do different jobs and are available in different kinds of planes flown by different kinds of pilots in different situations, so you really can’t compare auto-pilot miles to non-autopilot miles by just comparing deaths per mile.
- foepys 5y agoIn Germany 2,719 people died in traffic in 2020. In the US it was around 40,000. Germany has 1/4 the population, so maybe there are many other factors that can be tweaked before we let 5,000 lbs death machines roam around autonomously.