One of the things autonomous car fans have been fantasizing about is getting dropped off in front of (say) a store in a busy downtown area and having the car loop around the block until you’re ready to go so you don’t have to worry about parking. Or almost as bad, parking itself in a cheaper area farther away and coming back when summoned. A freaking nightmare scenario for traffic.
6 Likes
This 2014 NY Times article talks about how the rise of ride-sharing services like Uber and Lyft will turn cities into utopias. The author wrongly claimed that ride-sharing would reduce traffic, lower travel costs, increase housing supply, and even increase spending on public transit.
Despite apparent success of these ride-sharing businesses, none of these wild promises ever happened. Research shows that Uber and Lyft lead to a 40% increase in traffic congestion in the city. Uber makes cities worse. Lyft makes cities worse.
I mention this old NYT article because it is identical to the bold claims Waymo and Cruise are making about how their cars will improve cities. It’s the same bullshit wrapped up in a new layer of tech. We won’t fall for it again. We don’t need more cars on our streets.
2 Likes
Why not all the above? NYC has extensive public transportation and growing bike lane networks, but we still suffer from out of control vehicular violence. I am personally in favor of making free street parking illegal, expanding sidewalks, reducing travel lanes, but that won’t change the core bad behavior of drivers. Calling self driving cars a tech bro fantasy is stupidly dismissive and I know I should 't bother replying to boingers in general. But the fact is that autonomous mode within cities that would limit speeds to 20-25mph and prevent running red lights and other agressive behavior would be a major improvement on our quality of lives. Just because the poster boy for this tech is a piece of shit doesnt mean that the tech itself is a bad idea, or that it cant be implement in an intelligent and effective way.
Is that kind of speed reduction something that any of the autonomous car companies are proposing or pursuing? To my knowledge the cars run at the current speed limits.
It’s absolutely possible to design streets to be more pedestrian friendly and have lower speed limits, and that’s been successfully implemented in various cities and neighborhoods around the world, but that generally has nothing to do with whether or not the cars running on those streets are autonomous.
Also, you want to make free street parking illegal, which certainly would have some societal benefits, especially if it encourages use of public transit, but in an era of ubiquitous self-driving cars it would create incentives to have the cars keep circling rather than park, which makes traffic much worse.
7 Likes
If a machine is unable to understand traffic control personnel and unable to take direction from traffic control personnel, it’s not able to obey traffic laws and so should not be eligible for a driver’s license.
Part of the problem with self-driving cars is that we have very heterogeneous driving environments, and automation simply can’t keep up with them - until we have automated systems with real reasoning and understanding (and no, we don’t, we have various kinds of statistical modeling that could form the basis of a reasoning and understanding system), they are not going to be as safe as human drivers – and from 1940 until quite recently, the public argument for automated vehicles has been that they would be safer. (Now, of course, the public argument is for convenience, while the VC argument is for eliminating labor costs).
5 Likes
Anyone with even a rudimentary understanding of ML algorithms and their limits is terrified of autonomous cars. In a chaotic system like a city’s streets, there are always going to be phenomena too far outside the domain of the training set and there’s no telling what the resultant behavior will be.
6 Likes
Calling that stupidly dismissive is stupidly dismissive.
Since ride-share makes things worse, why does idealized ride share make that any better? No pie in the sky “they behave better”, actual research. One that involves all the sloppiness that would be inherent in people loading and unloading.
What you want is just as easily satisfied by mass transit only, no private vehicles.
6 Likes
Yep my level of understanding of the ML side and what goes wrong is more hobbyist, which is scary enough after years of “AI” fails… my professional understanding is data engineering and I know full well how poor quality a training set can be even before you put a human in front of it with their bias, trying to model new training sets to the problem you’re dealing with (cities are not uniform across the country), and just weirdness that has to exist for any data application that makes your code only as good as the most complex problem you have and how fast you can do it.
3 Likes
Really well said. I’ve done a lot of data cleaning in my day, and while I don’t work in the ML space, I’ve seen how badly poor normalization and other “dirty” data practices can affect at least some ML algorithms.
2 Likes
Self-driving cars that are capable of doing all the things that a safe and qualified human driver can do (like follow signals from a human being directing traffic) are, at least for the time being, a tech bro fantasy.
Most of the proposals you just listed that would actually make streets safer, like lowering the speed limit, don’t require self-driving technology anyway.
7 Likes
and yet one of the companies developing the technology deliberately courts aggressive behavior.
the regulations on all this are lacking. likely because if they were adequate none of this technology would be on our roads at all
The National Highway Traffic Safety Administration (NHTSA) said the Tesla software allows a vehicle to “exceed speed limits or travel through intersections in an unlawful or unpredictable manner increases the risk of a crash.”
3 Likes
I did work in ML for many years, and neither me nor any of my colleagues would go anywhere near the current generations of these cars. I maintain we’re still decades away from them living up to the vision people have for them.
Layfolk always see a tech that is 90% there and think it’s almost done. But as any software engineer in any field knows, the last 10% is 90% of the work. Some of those details, like “can’t recognize or follow a flag person” are deal breaking details that may not even be solvable. We don’t know yet.
5 Likes
/me Feels seen 
4 Likes
The fact that it seems trivial doesn’t make it so. The first obvious objection is that people with hearing aren’t the only people on the roads. That’s before you get to the more subtle issues with language, recognizing the sound source in a sea of vehicles and similar issues makes all of these things harder than they appear at first glance.
Adding a bunch of mobile empty vehicles will crush most dense cities under a sea of traffic.
This topic was automatically closed after 5 days. New replies are no longer allowed.
