FSD is used in one of these articles for Fully Self-Driving cars, sorry. As in the fantasy Jetson's version, but not flying.
There will be highway lanes for FSD trucks and cars, maybe just taking over existing HOV infrastructure, but they are at least a generation away from widespread use in cities. People will take over manual driving when they get to their destination if it is at all densely populated. The whole "pick me up at home, drop me off at work, go park yourself or work as a taxi for a while, then pick me up after work" will never happen.
Drum has been way off on this but he also expects strong AI soon. I think he said 2030.
I recall that there was another commenter in those threads who was working for one of the FSD companies. Can't remember the pseud though.
Waymo has been driving people around (parts of) Phoenix for a few years now. Without safety drivers even.
But how much experience do they need in Arizona before they can test with humans?
I like my cars how I like my women-as-a-class: always ten years away from full autonomy.
I think that joke warrants some side-eye.
Hopefully the road we don't go down is redesigning streets and laws so they work ideally for self-driving cars, at everyone else's expense. I'm not 100% sure we don't, though.
Can't comment for attribution on this...
Waymo is giving real driverless rides to the public in Phoenix and anybody who's walked or drived around SF can see there are 2 or 3 other companies that aren't that far behind. This isn't "full self driving" in the sense of L5 autonomy. It's "geofenced" to urban areas that are well mapped (including details like stop lights and bus lanes). The trucking approaches are similar but focus on well-known routes between hubs.
I wouldn't try to predict when L5 autonomy becomes a standard package on private cars. The price has to come down a lot. Either with steady progress on sensors or major improvements in non-LIDAR systems like Tesla's.
People are nervous about turning judgments like this over to a machine, but I predict that at some point in the coming decades, the equation will flip and we'll start to see human driving as a dangerous self-indulgence. Personally driving a car will come to be seen like going maskless, boycotting vaccination, or voting for Trump. No decent person will do it.
I don't think redesigning streets is on anybody's agenda. Believe it or not these companies are trying to get approval to operate, period, and don't have the power to transform our built environment. If redesigns were on the table, I would guess it would be neutral things like better signage and clearer lane markings. Or deploying transponders with road crews to broadcast traffic modifications (more sci-fi but accessible to any car computer).
Waymo cars are very cautious compared to humans. I've been stuck behind one and it kind of sucks. It's probably less of a problem on city streets than on 45mph roads in the suburbs.
I wonder how much this can be generalized beyond the US. The American landscape is uniquely suited to self-driving cars by virtue of being uniquely (some might say to the point of dystopianly) accommodating to cars. I can imagine that a country with wide highways, many areas which make walking illegal (etc) are useful for self-driving vehicles in a way that a pedestrian and bike-friendly city is not.
Then again my (Scandinavian) country has a pilot project for a self-driving bus. And I know of a similar pilot in Helsinki.
An interesting aspect in that article is the hostility shown to Waymo vehicles by a few ordinary Phoenicians. Sort of a Luddite impulse?
I've long thought that the prospect of Ford Pinto-style liability assessments as a jury response to the hand-wavy 'no we're actually saving lives' response when one of these things ends up killing someone is a barrier. Maybe not, though . . .
I don't think redesigning streets is on anybody's agenda. Believe it or not these companies are trying to get approval to operate, period, and don't have the power to transform our built environment.
When the money is there, companies like this have a solid track record of forcing the law to facilitate their market (viz. TNCs, though at least those firms are out of the game now).
I'm not thinking so much of changes to infrastructure as writing laws to penalize people for acting contrary to the software's assumptions.
I think 15.2 is right. The liability issue as as big a barrier as the technical challenges. People are such terrible drivers that FSD could be safer overall even if they still killed thousands of people a year. But FSD caused injuries and deaths would be the responsibility of the carmaker, and until that changes somehow it's hard to see how widespread roll-out is going to happen even if the tech gets a lot better.
An example of 13 is the cars will never exceed the speed limit, not even a little bit. Which means they are driving 5-10mph slower than most humans.
The liability thing is a big problem, and it doesn't play well with current insurance schemes. If a driver injures someone through negligence, their insurance covers it (overlooking underinsured driver scenarios for simplicity.) It's not clear what to do with a self-driving car. It seems that the reasonable thing to do would to fold it into existing insurance scheme -- it's a safer car overall -- except that if one isn't insuring against driver error the natural thing to do is hold the company liable for a bad product. The result is that none of the incentives favor developing self-driving cars.
20: There, at least, I'm not sure I get the insurmountable barrier. Right now, sure. But if they really are far safer overall, wouldn't it be relatively simple to revamp insurance regulation so where manufacturers do take responsibility for occasional glitches, and onboard data helps figure out if it was indeed a glitch? And financially feasible, as it would translate to lower insurance costs for drivers, meaning more ability to pay for the cars themselves.
I recall that there was another commenter in those threads who was working for one of the FSD companies. Can't remember the pseud though.
Not sure if you mean me but I own/started a company making software (part of the software stack, not the whole thing) for self-driving cars; whatcha want to know?
To the OP, nobody in the industry who knows anything about what they're talking about is surprised that you don't see Level 4 AVs (which is to say, what you're thinking of when you say "self-driving car") being sold today. The technical problem is as vastly hard as anything else that's ever been attempted at scale.
What you're likely to see in the near-term (leaving Tesla aside) is a lot more consolidation among the companies that are less well-capitalized (compared to, specifically, Waymo) as the big players try to figure out who will actually be able to compete. Meanwhile insofar as they're able Waymo is going to try and wait everybody out, incrementally increasing a rollout that relies much more heavily than they will ever admit on tele-operation. At the same time you'll see a lot of movement in sort of autonomous car-adjacent worlds, like sidewalk robots, factory robots, delivery vehicles. A lot of those will also rely heavily on tele-operation, and will also not admit it.
By about model year 2026 you'll see a lot more cars that have "Level 2" or "Level 3" systems somewhat akin to what Tesla has now (except working properly, ahem). A lot of this is going to be driven by regulation, particularly in the EU.
Tesla will continue to be a murderous wildcard until, I guess, they goad the government into shutting them down or shutting the entire industry down. It's an open question.
22 Yes it was you!
Don't really want to know anything in particular, just happy to see you in this thread.
Incidentally, while Uber ATG was actually a reasonably sophisticated (if occasionally catastrophically, unconscionably reckless) program, Lyft's internal self-driving efforts were never all that serious. Everybody's assuming that things are going to resolve to Waymo, Aurora as a strong #2, and then... who knows. Zoox already got acquired, Cruise is likely to start struggling with capitalization, Motional is sort of a funny also-ran but you never know.
22: Thanks. Did you use this same pseud when we were discussing this a few years ago? That might help me find what people were saying then.
My rental Toyota scared me briefly when it started braking well before I wanted to slow down as I came up on another car while driving with the cruise control. I had no idea it would do that.
24: Suppose Waymo, Tesla, and Aurora all pan out. Will the technology continue to disseminate for others to copy, or is it so complex and proprietary that we fall into a tight oligopoly?
25: I did, yeah.
26: right, yeah. There's going to be a lot more of that, and it's going to be increasingly annoying as a driver.
The most annoying thing is that you can't see the windshield fluid reservoir and I ran out at the worst time.
27: that's a great question, and it's a question that Waymo would very much like people to lack clarity on. My suspicion is that once one person has done it, it'll be basically doable to copy, so the big players are hoping to control the whole market before that happens. And really the timescales on this are looooooong. Waymo in particular does not have any pressure to produce anything genuinely commercially viable any time soon, so they're happy to let everybody else overpromise, underdeliver, and struggle to raise capital.
Sidewalk robots are going to need guns or something. I want to tip the little fuckers over on their side and I'm really nice in general.
31: figuring out how to deal with people who want to mess with autonomous systems is a genuinely huge problem, not just for sidewalk robots. For instance, if autonomous cars work well enough, they will _always_ yield to a pedestrian, right? So will pedestrians just start obliviously walking in front of them? What if the vehicle is _not_ in fact operating in autonomous mode?
I would like to thank Robot President for his leadership in the Robot Civil War.
27, 30 -- Won't the companies have to keep everything secret and proprietary to prevent master villains from hacking into all of Chicago's taxis from their under-volcano-lairs? Nothing can be shared. It's for our own good!
Here's my prediction: really truly seriously for-real fully self-driving cars will go the way of the Turing test. As a philosophical concept, people have decided it doesn't matter nearly as much as people once thought it did. As a thing or program people use, at best it's a toy or for replacing really shitty jobs that no one wanted to begin with, but there are worse applications as well. On the other hand, lots of things kind of like it are common and harmless, maybe even good occasionally.
No offense to Robot President, presumably a real human being who knows and cares about this stuff more than I do, but after reading the stuff about level 1-4 FSD and teleoperation, I'm not sure how much he'd actually disagree with me.
32: My favorite "messing with the killer robots" application was one where they claimed a car's autopilot that had been trained to recognize fake stop signs, fake lane markings, and presumably tunnels painted on the sides of mountains.
When the time comes, you won't need to avoid fake tunnels painted on the sides of mountains.
China is testing self-driving cars, and they are trying to get them working by the next Olympics for use in the Olympic park. If China does it then we are in a mano a mano fight with world historical impacts. /cough/
I am deeply suspicious of self-driving cars for many, many reasons. Chief among them is the demonstrated evidence that teams building important technology have major blind spots when it comes to race and safety. (E.g., FitBits and the like have been on the market for more than a decade and manufacturers still haven't clearly communicated how poorly their tools work on darker skin, much less fixed the problem, and now it's looking like autonomous vehicles are worse at "seeing" pedestrians who have darker skin.)
Even the technology being rolled out as "helpful" to existing drivers makes a boatload of assumptions. I've just gone through and disabled a number of the so-called features in my new car.
Among other things, I find auto-locking doors to be frightening, not reassuring. (My assumption is that it's much more likely that my car will be in an accident and someone will NEED to get in to rescue me, rather than some bad actor attempting to open the car door for nefarious reasons while I'm driving or stopped at a light.)
I'm also really irritated by the way the gas tank has set up, which has an elaborate series of rules regarding the location of the key fob and the lock/unlock setting of the car (especially difficult if you imagine having a passenger in the car who is holding on to the key fob, and the car's auto-lock settings turned on). In addition, the gas tank has a design that means you can't use any non-gas-pump nozzle to put gas in. So if I'm out of gas somewhere and trying to add some from a gas can I have to count on having a special piece of plastic that's going trick my gas tank into thinking it's a "real" nozzle? Not what I would call a feature!
Tl;dr: I'm biased against certain automotive technologies but grimly prepared for companies to simply start adopting it regardless of its costs or benefits.* Probably faster in the warehousing and the TDL field, where workers already have little power to resist.
*Lest you think I'm exaggerating, the UK Post Office is now being shown to have falsely convicted 39 workers, sending some to jail for YEARS, because their software was wrongly accusing the workers of theft and the PO wouldn't admit that it could be wrong.
when one of these things ends up killing someone
They've killed lots of people. I think as a society we've basically decided that we're going to run this experiment in the wild and we all know some people are going to get killed and oh well, we'll find a way to blame it on the drivers and keep going. I think I said this the last time it came up: I can't believe that we let these things on the road.
Remember icing? (I don't think it's become inexplicable hip again, this is the same generation as the first time.)
When our oldest was an infant we had a car with a remote starter that had stupid security settings. Once my wife was putting him in his seat, noticed the car was hot, and fished out the key from her bag to turn on the car. Instead of putting it back in the bag she tossed the key to the drivers seat to pick it up in a second. As soon as she closed the rear door the car locked itself, locking the baby in the car. As noted it was summer and hot enough to be uncomfortable in the car, but fortunately the car was now running the AC. Unfortunately the remote starter turns off the car after 10 minutes if you don't get in and insert the key. So she called the police, they called the fire department with an ambulance in case the baby overheated, they called a tow truck with a guy who has a slim jim for just such situations. He managed to jimmy the car door just a minute after the engine shut off, otherwise the firemen said they were going to smash a window. Baby slept through it all.
Now imagine the car driving away with the baby.
...just in time to take the baby outside the radius of total destruction of a nuclear strike. Waymoses.
I never knew people from Phoenix were called Phoenicians but I guess there isn't really any other option.
43 is horrifying. Glad everything turned out OK.
so many low tech relatively cheap things would make getting around sf easier, more pleasant & less environmentally burdensome than self driving cars ... this is just a tedious timeline we are stuck in. pretty much the only techno-car-control innovation we need/i would support is remotely adjusting the top speed of cars in line with the speed limit (+ lowering the speed limit). also camera tickets. but we can't even get those bc omg infringement on drivers'
liiiiibbbbbeeerrrrttttyyyy!!! instead we'll get "innovations" that further infringe-endanger pedestrians & bicyclists in service to umc & rich people who don't want to take the bus.
If anyone needs their day ruined, read "Fatal Distraction," the justly famous Gene Weingarten story in the Washington Post about parents who accidentally left their very young children in cars on hot, or even just warmi-ish days, with consequences implied by the title.
41: We won't blame the drivers if we can somehow blame the bicyclists/rollerskaters/perambulator-pushers. What were they doing outside without sensors to divert cars?
50: that story still haunts me years after I read it.
I was thinking a lot about that Fatal Distractions story in the context of police mixing up guns and tasers.
The argument here about insurance still haunts me years after I furiously stopped engaging.
52: Yup. Doug ain't kidding about ruining your day.
53: Yeah. Weingarten's accomplishment is making you understand the perfectly horrible screwups that you are capable of.
I've never read that story and I never will. Unless I get stuck on the toilet with nothing else to read.
Don't leave your baby near a toilet, it might explode.
43 is a nightmare. Adding in self-driving. Yikes.
I seriously wonder about the people who live in neighborhoods where there is no public transit who are too old to drive and have not moved to walkable places. They give up their license, and they are completely dependent on other people. I know they should all downsize, and it's bad land use, but if there were a way to have self driving cars keep the nonagenarians off the road, it would be great.
My probably-wrong sense of the whole thing is that we could greatly improve safety, traffic congestion, and fuel consumption on freeways by making every vehicle on there run in autopilot mode, and that the technology to do that is a lot easier than self-driving elsewhere.
The police take your license if you call them and say you don't know where you are and you are at a clearly marked intersection within a mile of your house.
I expect the big tech cos to be the first to set up "FSD-first" streets on their campuses. Either them or the charter cities people.
60: ubiquitous self driving cars are an absurd proposal to fix the problem you identify, and that is not a problem of any note in sf anyway & yet here we are, awash in waymos & lookalikes.
61: yes, that's called a train.
i'd like to get off the ride where we invent absurdly expensive suboptimal policy/tech "solutions" rather than implement the already existing policy/tech that works fine elsewhere.
61.2: No, it really isn't. You can't drive your train seat to and from the station, nor are there trains available on the vast majority of actually existing freeway routes. Even if you hate cars and hate freeways, why wouldn't it be appealing to reduce their impacts?
65: The availability of rail travel is not exogenous.
65: there is a gigaaaannnnnntic hand wavey move in your 61 that makes my train-and-a-pony future look eminently practical in comparison.
They've killed lots of people.
No, Tesla has killed lots of people. Most autonomous driving companies have killed 0 people and work really hard to keep it that way.
To be fair to Tesla, I don't think they've hit Pinto-levels yet.
Arguing for trains is just being contrarian and anachronistic. What we need is self-driving cars to move on some sort of predictable track, possibly harnessed together for energy efficiency. Maybe a schedule for convenience. Almost like a bus but longer.
When I was a baby in South Florida, my mom once accidentally locked the station wagon with me inside. She apparently found a police officer to jimmy the door open within minutes, but then he yelled at her because he (incorrectly) thought I had been stuck in the car the whole time she was grocery shopping. It did not help that I had been screaming my head off, but I was reportedly a very colicky infant. My car, thankfully, is impossible to lock with the key inside, nor do I have a baby to forget.
i once had a suitcase lock get stuck & the tsa people were amazing when i asked if they could please use their universal opener on it - they couldn't *but* they very proudly summoned their supervisor. she popped it up on a shelf near her head height, leaned in an ear and had that sucker open in seconds. also told me how many digits it had deviated so i knew the new combo. i of course asked how i could shower her with kudos via official channels - best service evah!!! - but sadly she demurred. her staff stood about beaming, this was clearly their favorite parlor trick. understandable!
One time, I locked my keys in my car and there was a garage nearby. So, the guy had a slim jim and he tried but he couldn't open it. This took like ten minutes. My wife finally asked if she could try and opened on the first attempt. And that was before she went to law school.
||
Bleg: This otherwise unobjectionable laptop has annoying preinstalled Mcafee stuff. Uninstall? Ignore? Other?
|>
68: Elaborate? I'm happy to be skeptical about self-driving cars, but I'm not getting the above-it-all dismissiveness.
Repeating what others have said, but adding the developing world: I think that in all likely futures the vast majority of vehicle travel will occur in cities which won't be remotely regimented enough to impose (outside some closed freeways) the consistency the software will require; and further, in many or most of those societies those rich enough for AVs will also be rich enough for human drivers capable of leaving the AV-safe bubbles; or indeed for uncrewed air taxis (which I'm guessing are technically close to possible already). I think closed, especially intercity, freeways will be the exception, where the combination of governance, tractability, and return to capital will permit at least AV trucks in at least high- and middle-income countries.
Related to 78 and 40, questions for Robot President on the data side:
To what extent do they need to be collected from sensors on vehicles? To what extent are they transferable between different countries/environments? Do you see any company/industry/country having a particular advantage in data collection?
We had a series of Pintos and none of them ever killed us. Stranded us, sure.
Since there's two anonymous empty strings in the thread, I'm the one through 69. 76/77 is somebody else.
76: Uninstall the Mcafee stuff, and just use Windows Defender, or whatever it's called these days.
The American landscape is uniquely suited to self-driving cars
True, but I think ubiquitous AVs are far more likely in China (and indeed unlikely anywhere else, excepting maybe SG, some GCC, maybe ROK). TL;DR: the US political economy will tend to push software to improve until it's good enough to handle complex reality; the PRC political economy will tend to bury complexity in concrete until reality is simple enough for the software to handle.
Many things that are already central to PRC political economy and policy can serve also to make AV adoption dramatically easier: saturation data collection; constant rebuilding of infrastructure; compulsory trading up of vehicles;* demolition of large swathes of organic built environment; regimentation of large swathes of organic urban activity; suppression of opposition among those adversely affected by the foregoing. And, if successfully executed, AV ubiquity would potentially release a huge quantity of labor productivity.
Explicitly: every congested street turned into a wide-open access-controlled boulevard; every streetfront business consolidated into an enterprise big enough for a party cell, and moved inside an access-controlled compound; every person traveling only in predictable remotely-controllable vehicles, walking only through access-controlled grade-separated walkways. This not just something the PRC could do, but would want to do and is in many ways doing anyway.
*In that they push EVs and I assume fuel economy and pollution standards; IDK to what extent that applies to replacement vehicles rather than new ones.
every streetfront business consolidated into an enterprise big enough for a party cell
Here we call that a multilevel marketing scheme.
77, 83: well, what mc said in 84 ... actually, one thing that brings a smile to my cynical mug is seeing the waymos paralyzed by grandparent shopping hour at like jackson and stockton.
the safer, more fuel-economy-ey*, less-congested** freeway of the auto pilot future skips over how we get to every car that needs to get somewhere via freeway having autopilot. a hell of a lot of poor people depend on really old cars to get to their jobs. so once we've subsidized these folks getting auto pilot retrofitted in their old cars why not just build the train system? or, alternatively, they can't use the freeways with their crappy old cars to get to their jobs. great result.
then all the auto pilot cars get on the freeway - to achieve this utopian freely flowing traffic the freeway is going to need retrofitting. checking that all the cars have auto pilot, monitoring that it stays on, etc etc etc. again, why not just build the train system? use the freeway row!
*color me skep-ti-cal! and don't waive ev at me - an enormously heavy electric suv with a small number of people in it (let's be honest - just one) is not fuel efficient in the scheme of fuel efficiency we need and that is demonstrably achievable by a rational mass transportation system. and that extra weight comes with air quality and road maintenance issues/costs.
**hahahahaha. i've yet to meet a "solution" to single vehicle congestion that was effective and durable. and i've absolutely served my time in the trenches with traffic engineers.
OT: Guard your science teachers. There's another divorced billionaire looking.
why not just build the train system? use the freeway row!
Isn't one answer that trains are less efficient the more stops they make, whereas a freeway of self driving cars allows people to pull off at any exit (also, if freeway upgrades are required they only need to be done in urban areas where traffic is densest-- you could keep the existing freeways between cities, whereas building train systems requires replacing the whole thing at once).
I would _love_ to get rid of most car traffic, but I think upgrading cars is easier than building a full fledged replacement (particularly at current US population density).
I'm not against trains, just willing to back EVs if they look like they will work. Here's David Roberts on the benefits of promoting EVs: https://www.vox.com/energy-and-environment/2019/8/12/20801602/colorado-electric-vehicles-2019-renewable-energy
And this has a section on how to help people afford the transition (low interest loans): https://www.vox.com/energy-and-environment/21349200/climate-change-fossil-fuels-rewiring-america-electrify
79: I didn't really understand the question, but if this is an applicable answer the question of how much real world driving data can be augmented with simulated data is very much an open one. The answer seems to be that it depends on what you're simulating; interactions with other vehicles are going to be easier than interactions with pedestrians, for instance. To the question of how much you will need to retrain for new environments, the answer is probably a lot. Small differences in the built environment produce vastly different driving behavior, and that's leaving cultural differences aside. You'll note that the vast majority of current AV deployments are places -- suburban Arizona -- where driving is notably "easy".
Incidentally there is no particular reason for me to pick a side on this, but I think those above who say "if you are going to spend a ton of money building special infrastructure, why not just build trains" have an excellent point. Freeways also work less well the more exits they have. From my perspective, I regard the challenge of self-driving cars (which is, to reiterate, almost unimaginably massive however you slice it) building vehicles that can operate in the world as it exists. If you are building both vehicles and the infrastructure for them to operate in, why would your solution be a ton of expensive compute and fiddle computer vision when you could achieve the same effect with simple technology like rails and centralized switching?
89: I think there's a significant gap between "EVs are genuinely beneficial as a climate change wedge" and "autonomous cars + smart freeways are a good substitute for Moar Trains".
90: Thanks. I was getting at the extent to which data from static sensors like surveillance cameras can supplement data generated aboard vehicles.
(I'd also really appreciate an email at the linked address.)
Oh, I see. Well, it relates to my answer, in that you can potentially use data from fixed sensors to build simulation environments/scenarios, but insofar as what you need to do is train computer vision and/or inference from other sensors, that's definitely highly viewpoint dependent (enough so that, for instance, data from a camera mounted above a truck's windshield is not generally useful for training models for cars, because of the viewpoint difference).
"Self-driving" probably includes significant remote human control as noted above, but that eliminates what's being sold as a big benefit of autonomous driving, the ability to pack more cars in existing space or eliminate some traffic control systems and instead rely on vehicle to vehicle comms. If anything remote control goes in the other direction, since reaction times are longer for a remote connection.
All I want is the ability to hop into a car with a case of Hamm's, some ice, and a bucket of fried chicken and get to Philadelphia safely in five hours. Is that too much to ask?
Isn't one answer that trains are less efficient the more stops they make, whereas a freeway of self driving cars allows people to pull off at any exit (also, if freeway upgrades are required they only need to be done in urban areas where traffic is densest-- you could keep the existing freeways between cities, whereas building train systems requires replacing the whole thing at once).
If you're talking about energy efficiency, the answer is definitely transit. The IPCC modeled and found that widespread adoption of EVs is unlikely to be enough to let us meet carbon goals; it has to be combined with less driving and more transit period.
What gets you between train stations is buses.
Urban freeways should be destroyed altogether, they're the worst use of urban space after surface parking.
Those are up there but they aren't as carbon-intensive! #colmaforlivingpeople
89.2: faced with two options - 1) spread the cost of transition to a less-energy intensive, non-carbon future to the entire group of taxpayers where we collectively decide how to split the burdens, or 2) saddle low-income individuals with loans (presumably secured, i.e., you lose the car if you default, and my god possibly personal recourse we live in a sick enough society), i find the choice easy. ym, as they say, mv.
~ducks~
(One factor is we both can't and shouldn't replace existing cars with EVs much faster than our current cars die off, because of the carbon cost of making a car is high and offsets a lot of benefit. All that steel!)
We'll still need the cemeteries for the golfers. Are you the kind of monster who would just cremate them?
100: I agree completely, that if the costs and ease of adoption are comparable that it's an easy choice. I'm not sure that they are, but I can see what I can find to support that belief.
102: How true is that? I've heard that, but here's David Roberts (who I tend to trust) arguing the opposite: https://www.vox.com/2015/11/16/9737720/electric-vehicles-cleaner
Here's something that takes the pro-train and bus position, but I'm not able to read it to see what are the comparative costs. Does anyone have access: https://www.nature.com/articles/s41558-020-00921-7
Export the used vehicles. Many (most?) countries are adding vehicles faster than electricity supply. And even if scrapped outright, scrap makes for much cleaner steel than ore does.
Anecdotally, I see a huge number of used cars being towed individually to Mexico.
You could put a bunch on a car carrier and reduce emissions.
There's apparently a huge used car thing going on right now in the US. Inventory way down, prices way up, and it's affecting car rentals. Pandemic related, so it won't last, but it's also a limited universe, so impacts last.
109: we should build some sort of track.
Skepticism for rail replacing highways (between cities), while supporting local transit: https://noahpinion.substack.com/p/i-dont-get-the-high-speed-rail-thing
It cost over $100 a day to rent a car in Omaha. Bastards.
did i say one everlovin' thing about high speed rail? no i did not. i live in the inner bay area. we have existing bus, light rail, commuter rail and interurban rail systems. robust investment in them would reap massive beneficial returns across multiple sectors and really change ordinary peoples' live for the better. self driving cars solve no problem in the actual place where i live. they are gigantic tedious distraction.
114: I was thinking about what it would take to completely replace freeways, but I may have misread you. I 100% agree with scaling up local transit.
The reason I started down that path was that I wasn't sure you could fit the existing freeway and trains in the same right-of-way. But, again, I think that sent me on a tangent.
Based on my experience, most people who rent a car, rent it in Omaha, so it's a huge problem.
The remote capabilities of autonomous vehicles are not remote control. The car can send back simple multiple-choice questions to a human, like "which way is this crossing guard pointing?" The car cannot be driven remotely, period.
I just don't understand how you can rent a car in Omaha for $100, sell it in Malta for $0.03, and make a profit.
Oh, so that's why I'm asked to identity all the squares with a street sign in them.
Yep, city and regional rail and buses are the low-hanging fruit. HSR should first be between close regions (SF-LA, Dallas-Houston, etc.) to crowd out air/highway, then lastly connecting the country up. Although the federal government could do this all at once too.
105: The numbers there compare carbon over the whole life cycle; it doesn't seem to dispute that EVs take a ton of carbon up front, although it says that's coming down too. It's disputing the proposition "EVs are actually worse overall", whereas mine was "we should replace gas cars with EVs only as gas cars die, not try to goose that to 5-10 years total replacement with more subsidies [than we have now]."
112: Noah really buries the lede there: "This is not to say that HSR has no benefits over cars and airplanes; the benefits come in terms of convenience and climate impact rather than actual cost."
i would absolutely take an overnight sleeper train back & forth on business trips to LA, doesn't require high speed, but alas even that is beyond us. pre pandemic i'd found an overnight sleeper bus service, hopefully that has survived & can demonstrate good ventilation. if we can't have ordinary plain vanilla train service, this would be a good candidate for evs. a silly kludge solution, but oh well.
DQ, we have pedestrians and bicyclists getting killed and injured by human drivers all the time, but go off.
123: yes and we know exactly what we need to do to eliminate those deaths, self driving cars not part of that toolkit. chose yourself how you want to spend your time and energy, i'll do the same.
120: Apparently the Duolingo guy and the original captcha guy (ie read scraps of old books for the archivists) are one and the same.
They are "not part of the toolkit"? Good to know. Thanks for sharing.
||
Does depression interfere with memory formation?
|>
honestly 127 unless you are jeffrey tumlin - fuck off. and if you are jeffrey tumlin commenting anonymously on this obscure benighted website, triple fuck off. in sf we know exactly where & how fatalities occur & what we need to do to prevent them we just need the political will to do it. plus note in my first comment my endorsement of the tech of limiting the speed at which all vehicles travel and lowering the speed limit.
No, Tesla has killed lots of people
Fair.
There is only one person in this thread who needs to fuck off. See if you can guess who.
NickS, a while back you mentioned a book about 2nd hand stuff. Did it talk about vehicles/engines?
NickS, a while back you mentioned a book about 2nd hand stuff. Did it talk about vehicles/engines?
No, I don't think so. A mention of scrap metal, but not used vehicles -- in part because it was focused on things that might me less visible (and because system D makes for good stories).
136: Thanks. (I was thinking even ripping out ICEs and shipping them to the poor world as generators might make for net carbon savings given fuel economy in newer vehicles. Which if true is presumably already being done.)
Maybe the developing world would like a bunch of giant pickups that were $60k when new. Because that's what we have now.
Would it be practical to have self-driving buses or minivans on very short high frequency high density routes? So everyone is within 1/4 mile of an automated pick up, which feed into arterial buses, which feed into rapid transit. It's more transfers which are bad for usage but with enough frequency (
<5 min on the short routes and arterial buses) it's not much worse than a taxi direct to trains. And because the local routes are short and repetitive the automation should be easier to adapting.
A few thoughts:
(1) I'm fairly anti-car. I dislike driving and generally commute by bike & bus. For a couple years I was working in an adjacent city one day a week and commuted by bus (~1hr 30 each way, which included walking to the bus station, and a 10 minute wait for a transfer which gave me a chance to stretch my legs). At the time I was surprised at how well that worked. Eventually I inherited a car and was surprised at how much better it was to drive 45min each way -- even with my dislike of driving. So I'm very pro-transit but also understand why it's likely to replace some trips but not all of them.
(2) I was skeptical of EV adoption for a while -- I liked the idea, but didn't see it getting enough market share to matter. Now I'm moving the other direction; I think there's a chance that EVs could take off, and I think it's worth pushing hard in that direction. Not because I think private cars are great, but because I think it would make a big difference and it might be achievable.
(3) When it comes to fatalities, car accidents kill, what, 38K people per year, and there's an estimate that air pollution (only some of which is from automobiles) kills 250K/year. There's a good argument that replacing IC cars with EVs or transit (whatever works) will save as many lives as increased safety measures.
was skeptical of EV adoption for a while -- I liked the idea, but didn't see it getting enough market share to matter. Now I'm moving the other direction; I think there's a chance that EVs could take off, and I think it's worth pushing hard in that direction. Not because I think private cars are great, but because I think it would make a big difference and it might be achievable.
EVs or AVs? EV adoption seems pretty much inevitable, given that most countries are banning new IC engines in the next 10-20 years.
I suppose I shouldn't say most countries, but certainly countries with most of the world's car sales (though China's ban is a bit of a fudge).
Oops, can that be pseudified, please?
Yeah, one can imagine everyone toodling around in their electric F150s and Silverados in 2030 but autonomous driving still being restricted to, like, warehouse forklifts.
Those are too small. Real men have bigger trucks.
For a couple years I was working in an adjacent city one day a week and commuted by bus (~1hr 30 each way, which included walking to the bus station, and a 10 minute wait for a transfer which gave me a chance to stretch my legs). At the time I was surprised at how well that worked. Eventually I inherited a car and was surprised at how much better it was to drive 45min each way -- even with my dislike of driving.
Hypothesis: you're an agreeable person who sees the good parts of situations.
139, 140: in theory, although unless you can keep speeds under 10mph or so the engineering challenge is not that much easier (since you still have to interact with humans).
When it comes to fatalities, car accidents kill, what, 38K people per year, and there's an estimate that air pollution (only some of which is from automobiles) kills 250K/year. There's a good argument that replacing IC cars with EVs or transit (whatever works) will save as many lives as increased safety measures.
That argument is valid, but one qualifier is that a lot of the air pollution cars cause is from tires.
|| I need a ruling from the lawyers. I file a motion to dismiss a complaint for lack of pj. Other guy calls me, says he's slammed with work, asks for a 2 week extension. 10 days short of his new deadline, he files an amended complaint, which adds claims and (allegedly non-diverse) parties, but does not address the pj issues I raised. I'm kind of inclined to tell the guy not to ask me again for an extension based on workflow when he's going to be doing extraneous shit in my case. Unreasonable? |>
I file a motion to dismiss a complaint for lack of pj
I know standards of attire have slipped in the age of videoconferencing, but I didn't think it had got quite that far.
151: Not unreasonable. He misprepresents his reasons for needing professional courtesy, why should he get it in future?
149 cont'd: there are some projects, like Toyota's "Woven City" concept, that are looking at transformed urban spaces where you have low speed autonomous shuttles and delivery vehicles, 100% shared spaces, and lots of multimode movement (i.e. bike, scooter, ped, etc) with zero traditional autos. Something like that does give you a somewhat easier (engineering-wise) way to get to a last mile solution like SP proposes, but it requires basically (or, in the case of the Toyota project, literally) building entirely new city centers.
Our existing city center looks really expensive to redo.
Amsterdam did it over a generation by adding new standards all streets had to follow as they came up for refurbishment. I assume it was relatively expensive overall, but a lot of it was just repurposing streets for non-car uses, so it wasn't rebuilding the city.
I'm assured by the Saudi government that all you need is one long tunnel down the middle of your city.
Did that work for Boston? Honest question as I have no idea.
157: yes, it's somewhat more of a political challenge than an infrastructural one in older cities. Different deal in, say, Phoenix.
The low speed autonomous shuttles thing, slightly appeals to me, because it has a 1970s sci-fi movie vibe about it. But ... what problem is being solved in a lot of these use cases, that isn't solved equally well by: walking or cycling or electric mobility vehicles for people with mobility issues? If I was going to engineer city streets to make them more usable, there's a lot of things I'd want to improve first, for human users. There aren't even usable pavements in a lot of where I live.* The nearest bus stop to me is about 400 metres away, and the nearest tube station is about a mile. Most people don't need any kind of transit to take them closer than that.
* largely because they are parked up by car users so tightly that you can't get between cars and walls, or because they 100 years old, narrow and have huge stone kerbs which are not friendly to wheelchair users or people with kids in buggies/strollers.
I know David Roberts was really excited about Barcelona. I admit, I haven't read the whole series.
The challenge when implementing superblocks, then, is to preserve adequate circulation of people, goods, and services using fewer vehicles on fewer streets. As discussed in part 3, that will mean moving people into other means of getting around, like buses, bikes, and feet. And it will mean channeling the vehicle traffic that remains -- buses and (hopefully someday electric, shared) cars -- onto fewer orthogonal through routes.
"From the street grid in the city, we select those streets that maintain the conditions for transversal mobility," says Ton Salvadó, Barcelona's chief architect, "and the rest of the streets are calmed."
They've implemented LTNs (low traffic neighbourhoods) near me, which involves closing all the through routes across the borough and forcing drivers to skirt the periphery. I think it's fair to say that public opinion is overwhelmingly hostile. There have also been a number of speciously reasoned articles in favour of them in the Guardian.
If you would like to come up with a positive use case for autonomous vehicles, low speed shared spaces are a pretty good one, as some users/uses do still require higher capacity motorized vehicles (at least for a reasonably high quality of life while out and about) and autonomous vehicles, unlike humans, never get frustrated or impatient.
163: Yes, I think you've mentioned that the remaining arterials are inconvenient and that it makes everyone's driving experience much worse.
I'd hope there are ways to minimize that problem, because there is something really attractive about the suberblocks idea.
The nearest bus stop to me is about 400 metres away, and the nearest tube station is about a mile. Most people don't need any kind of transit to take them closer than that.
I think heat exhaustion will be imposing some revisions there. (Not that I think AVs will often be the best solution.)
re: 165
Yeah. I live next to one of those arterials and the traffic is so dense that I couldn't safely cycle* up the road the other day. I had to go off road through a park to get home.
The net effect is that traffic has been funnelled onto roads that are primarily multiple occupancy (flats, or shared housing) or low income housing away from more "leafy" streets. My son's school now has a traffic jam outside it for about 3 hours a day. The problem is that they've not done the work they'd need to keep traffic flowing where those arterial roads meet, so traffic is being funnelled into junctions that simply can't cope with the traffic flow and then local people left to live with the consequences.
I think the explicit plan is to make driving so intensely unpleasant that people stop. I'm not even entirely against that as a strategy, but they need to combine it with more "carrot" as well as stick to make alternative modes of transport attractive. The net effect, though, has been to make those roads that are not part of the low traffic scheme much more unpleasant to cycle or walk along, without any corresponding gains elsewhere.
* not because I'm either a timid or a slow cyclist. There literally wasn't room between the stationary cars (in the traffic jam) and the cars parked at the side of the road to pass through on a bike, and the oncoming traffic on the other side was so fast and dense that it wasn't possible to safely filter along to the outside of the traffic jam, either.
I'm assured by the Saudi government that all you need is one long tunnel down the middle of your city country.
Arguably the Big Dig has had a positive ROI on the $14B whatever. Obviously hard to disentangle other growth- would they have built a new arena? New apartment buildings? As much biotech and finance space? But you can make a pretty direct claim that the entire new district east of downtown wouldn't have happened if there were still an 8 lane highway cutting it off from the rest of the city. A district which, incidentally, contains the area where they test AVs in Boston.
The idea of low speed autonomous vehicles in urban areas (I guess as delivery trucks?) seems weird to me. Why not employ a few drivers? The lower speeds would make it a more pleasant job, and the fact that the vehicles would be interacting with lots of pedestrians, cyclists, etc. suggests that real people should be driving them.
If there's a driver, you can really throw rocks at the vehicle without being a monster.
169: in practice (from personal experience) driving at under 10mph in an environment where you are constantly required to yield to pedestrians and cyclists and whatnot entering your path is pretty maddening, at least when you have somewhere to be. Which isn't precisely a counter-argument -- people can learn to do maddening things with equanimity for a job -- but the idea is that you lose the risk of emotion-induced bad human behavior while gaining the benefit of a vehicle that can and will stop instantly for anything in its path, which in certain limited circumstances can be beneficial.
I've found coming to a complete stop at stop signs (not the norm here) a useful exercise in ego-death. YMMV.
If one simply builds low rise on a fractal grid superblocks kinda just happen. Admittedly with default private vehicles being scooters not cars.
Maybe FSD cars should be made out of fine porcelain and tissue paper. Travelling Faberge eggs.
Anecdotally, I see a huge number of used cars being towed individually to Mexico.
In Belize, I met a guy who once a year would get himself up to central Mexico, buy a used car, drive it over various godawful roads to get home, and sell the car at a nice premium on the local market.
town car dude your super content rich penis having contributions to this thread have clearly vastly outweighed mine, super great to witness you perpetuating the hallowed traditions of unfogged.com, particularly noble to do so presidentially. bravo everyone.
hi bave!
It's not like anyone doesn't know who he is, right?
I actually don't have a strong guess. Certainly a jerk, but I couldn't say who I thought it was.
Lurkers in email are telling me I'm wrong.
I will note that whoever Abe Lincoln Town Car is (looking at the IP address was not illuminating), they're doing presidentiality wrong. It's to allow for revealing personally embarrassing information, not to be ruder than you would feel comfortable being under a persistent identity.
I feel like when I assumed everyone knew Lindsey Graham was out.
I'm the one being rude?! I invite you to review the thread.
I'm presidential for the same reason Robot President is, which I thought was clear from 9. I've commented on this blog for like 15 years but it's about time to stop I think.
I was 9, 11, 13, 19, et al, BTW. I didn't come in here hot lecturing everybody who disagreed with me.
2nd 187. Although I broadly agree with DQ on the merits, crucially:
we invent absurdly expensive suboptimal policy/tech "solutions" rather than implement the already existing policy/tech that works fine elsewhere.
186, 187: I think not having signed your first several comments at all made it unlikely for people to have connected them to each other -- looking back you did, I guess, claim all the unsigned comments except for two, but I certainly didn't go back through the thread looking for unsigned comments when you did that, and I doubt anyone else did.
Now that I understand why you're anonymous, that's fine, but I'm still not really following what points you mean to be making generally.
174: Locally, I feel like I'm at risk stopping at the newly installed, but not yet turned on stoplights. I know the DMV says to treat them like stop signs, but I worry about being rear-ended by the drivers who ignore them until they're operational.
Is there any chance the robot cars will take over the streets in the next couple of weeks? Asking for a friend who has a kid who wants to get a learners permit.
192: you live in robot car country. Go knock on some doors and ask them to hurry it up.
I threw too many rocks at their cars.
I'm still not really following what points you mean to be making generally.
That's on me for not keeping up the threading. I was replying on my phone.
To belabor the part where all this got contentious, in 114 DQ said, "self driving cars solve no problem in the actual place where i live." My response in 123 was that they could improve safety for pedestrians and cyclists. Pedestrian and cyclist safety is very much a problem in the actual place where (to my understanding) we both live. And AVs will improve safety over human drivers. They are cautious to a fault. (See also 13 and 19.) DQ responded, "self driving cars not part of that toolkit," as if it is impossible that they could contribute in any way. I found that obnoxiously dismissive and responded in kind.
I was probably overly snarky starting in 123, but I have found DQ's comments in this thread quite grating--particularly their lack of any epistemic humility--and I let my annoyance show. To have 129 and 179 result in me being called out as a jerk is... fuck, man, I don't know.
190: With all respect and affection for DQ, I don't understand the level of hostility in this thread. To the extent I understand her position at all, it's that human decency requires that all cars be replaced by trains and buses and that any discussion of marginal improvements to car-based transportation is evil and wrong. Which is not to say that I actually think she thinks that, but just that I'm mystified by the hostility.
Not to speak for DQ, but I understand her point of view pretty clearly -- that a combination of mass transit and traffic calming gets essentially all the social benefits, in safety and utility, people are hoping for from autonomous vehicles, and doesn't require any technological advances. This seems obviously true to me, depending only on how much you value people being able to travel in individual vehicles.
On the autonomous vehicle side, I'm seeing people hoping that technology will advance to the point that they're as useful as and less dangerous than human drivers, but they certainly aren't there yet, and it's uncertain whether they will be anytime soon. And there's some risk that adopting them before the technology is ripe will result in significant risks or loss of utility to the people who are involuntarily forced to interact with them.
Lifting my gaze from my navel, my case for AVs:
1. They are safer, for riders and non-riders.
2. Being safe, cheap and convenient, AV taxis will displace private vehicle ownership in dense urban cores. This will:
(a) Allow increased density due to decreased demand for parking.
(b) Accelerate electrification as trips shift from gas cars to AVs.
I don't think they will completely displace mass transit and I also don't think they will somehow fuck over marginalized groups. They will encourage continued investment in passenger car infrastructure which could slow down necessary drastic shifts towards greener alternatives. But car culture in the US is so far beyond untouchable that it's hard to imagine the marginal effect here is more than a drop in the ocean. If autonomous cars and tracks pan out, autonomous buses should be on the road before long.
On the "where'd all the hostility come from" point -- DQ's comments before 120 are very hostile to autonomous cars, but aren't arguing with or insulting any particular person in the thread (or any specific person at all). Abraham Lincoln Town Car got directly but kind of cryptically disagreeable with her in 123, she snipped back, and then it was off to the races.
On the autonomous vehicle side, I'm seeing people hoping that technology will advance to the point that they're as useful as and less dangerous than human drivers, but they certainly aren't there yet, and it's uncertain whether they will be anytime soon. And there's some risk that adopting them before the technology is ripe will result in significant risks or loss of utility to the people who are involuntarily forced to interact with them.
There are also the opportunity costs and the effects on other decisions. This seems to be happening less now that the limits of current technology are becoming clearer, but for a while a few years ago there was a lot of stuff like transit agencies saying "well, we could significantly upgrade service in our next ten-year plan, but then again it's very likely that self-driving cars will replace all other transportation modes in the next five years so it would be easier and cheaper to do nothing instead." That sort of thing is I think what DQ is mostly responding to.
198: When you say "are safer" you mean "hopefully will be safer when the technology improves in ways we don't yet quite know how to do" right? There isn't an AV in existence that can drive from my front door in NYC to anyplace safely, as far as I know.
I do like to toss out the counterfactual to people -- it has more or less been mooted by DQ in this thread, I think -- that it would be a much more doable engineering task to add the safety features which will come automatically with autonomous vehicles to existing road vehicles; you could pretty straightforwardly add features that would not allow speeding, and which would enforce stopping completely for red lights and stop signs. You could do that today, at quite high level of performance. That's in addition to features which already exist, like automatic emergency braking. It's very likely that would be sufficient to almost completely eliminate road deaths, or at least road deaths caused by cars with those functions installed and enabled.
Nobody's going to do that, of course -- it would be as widely loathed a development in the history of transportation as I can imagine -- and I actually think the at least some of the reasons nobody would do that are pretty valid. But I'm not sure there's a good argument that what I said isn't true.
That really is interesting. The idea of it being possible to involuntarily enforce speed limits for all cars on the road would change driving incredibly.
a combination of mass transit and traffic calming gets essentially all the social benefits, in safety and utility, people are hoping for from autonomous vehicles, and doesn't require any technological advances.
I'm in the both-and camp here.
On the autonomous vehicle side, I'm seeing people hoping that technology will advance to the point that they're as useful as and less dangerous than human drivers, but they certainly aren't there yet, and it's uncertain whether they will be anytime soon.
Here we just disagree.
202.1 is pretty much where I was going way back in the thread with a comment about autopilots on the freeway, which is a really easy driving situation that humans nonetheless screw up massively due to various combinations of aggression, inattention, and plain old incompetence. I may be missing something important (how the autopilot responds to a plastic bag blowing across the freeway?), but the basics of safely merging into traffic, tracking down a lane, and maintaining a safe distance from surrounding vehicles don't strike my unenlightened eye as obviously more difficult or expensive than other technologies that have become ubiquitous in the past 20 years or so. Pursuing those improvements while also improving transit and reducing car reliance doesn't require going full Elon Musk.
204.2: Do you disagree about whether we're there (AV can do anything a human driver can do, more safely) yet? Or just about whether it's definitely going to happen soon?
201: looking it up, here's the information Waymo has released about safety. I'm not sure how it compares to human drivers (and is the fair comparison to an average driver or one who is as cautious as the Waymo vehicle?) But it's not obviously unsafe: https://www.theverge.com/2020/10/30/21538999/waymo-self-driving-car-data-miles-crashes-phoenix-google
I have no particular forecast about AVs, but the point that Kevin Drum, for example, has made is that _if the technology is sound _ the time it will take to get from, "working fairly well" to, "clearly safer than a human driver" will be short. That's a big "if" however.
201: They are safer where they are able to operate, e.g., Phoenix and the SF Bay Area. I have no idea when AVs will be able to operate freely in Manhattan. I think there's a big gap in appreciating how advanced the technology has gotten, particularly in the past year, for people who aren't in areas where they operate.
There are fully autonomous vehicles operating driverless throughout the Bay Area? If that's true, I had no idea.
I have to say that I don't believe that to be true -- I think you must be saying something much more limited.
it would be as widely loathed a development in the history of transportation as I can imagine
The closest parallel I can think of is in the early 1970s when the government tried to require an interlock system to prevent cars from starting unless the seat belt was connected, which caused such a backlash that Congress immediately backtracked and eliminated the requirement. (I had thought they actually banned automakers from including an interlock, but from that article it appears they just repealed the earlier mandate. Also, airbags: very controversial at the time!)
My dad buckled the belt and then cut it off to stop the buzzing. He felt that as the driver, it should be more risky for him than for his family. Since he was smoking with the windows closed, he couldn't use the seat belt.
Anyway, the thing that finally convinced me to wear a seat belt was breaking the windshield with my head.
209: No, but multiple companies have cars with safety backup drivers out every day. I see a Waymo or Cruise car nearly every time I leave my house. You only have to extrapolate from the Phoenix example to see where things are going.
They are safer, for riders and non-riders.
I admit to having followed the news on this only somewhat, but I thought this was demonstrably untrue. That is, the injury/fatality rate for AVs, per million miles driven, is way higher than for regular drivers. (Bearing in mind that they haven't driven many millions of miles, so a handful of incidents have a big impact on the rate.)
Am I wrong? I'm willing to read links that say I'm wrong, as long as they are reasonably well supported by evidence.
Also, I only visit SF, but even based on just three or four visits there in the past few years, I'm sympathetic to DQ's level of fury. If only because it is so unbelievably tiresome to deal with the level of techno-utopianism and disdain for public transit and the people who use it that is exhibited by some of the loudest voices in town.
There's a reason Twitter is full of jokes about how techbros reinvented the bus. And those jokes are punching up, not down.
25: there was this.
(I'm not so extremely online as once before but do check here every now and then.)
220: All I can say about that thread is I was not consciously aware I have been so consistent on or passionate about this subject for so long. It would have made certain developments in my career less surprising. Also, Tweety (who had a retroactive pseud change?) was fun to argue with.
218: 207 may help? I don't think there's any independent analysis, except maybe what regulatory agencies have compiled.
I know this guy 'cause of blogging and I am so so impressed and proud. I cannot wait to go to a show.
In re 215: "cars with safety backup drivers" is really, obviously, by definition not equal to 209's "fully autonomous vehicles operating driverless," and implying that it is makes me kinda hostile, and I'm not even in the States much these days, much less San Francisco.
Emitting vast amounts of handwavium while insisting that any year now driverless cars are going to solve myriad social ills (except for giving everyone a pony) seems to be the standard industry response, and it has -- as teo pointed out -- already had pernicious effects on budget and investment decisions. The record of highly visible companies over-promising and under-delivering certainly adds to my skepticism.
Add to that the amount of "Oh don't worry" from big tech companies when in fact people were quite right to worry and apply it to thousands of pounds of hurtling steel, and skepticism shades easily into suspicion. I see software glitches all the time, but at least I can usually reboot my computer; glitch a car and there's no ctrl-alt-del to restart my life. As someone who was a bicycle commuter and probably would be again, the blitheness of autonomous cars is a deadly peril that, given the track record of the car industry, companies will probably do their best to ignore. Given the history of bending the built landscape to accommodate cars rather than people, looking at autonomous cars askance is the very least one should do.
Speaking as a former cycle commuter, I would trust an autonomous car more than many of the autonomous human drivers of London.
225: detecting cyclists is a very challenging vision task; there is a reasonable chance that either you'll get detected as a pedestrian (and thus your anticipated movement profile -- used in planning the motion of the vehicle -- will not match your actual potential movements) or a somewhat smaller chance you won't get detected at all. This is a more-solved problem than it was a couple years ago -- I expect Waymo vehicles have very few false negatives with cyclists, these days -- but it's still a problem.
I'm afraid assholes will set their car to circle the block all day to save $15 in parking fees and traffic will be nuts.
detecting cyclists is a very challenging vision task
Why?
Come to think of it, I've never had a capcha that asked me to spot the cyclist, though I must have identified hundreds of traffic lights.
228: bicycles are mostly see-through and in general the angle that a cyclist presents to the sensors on an AV is an oblique one, this means that the actual bicycle is represented by relatively few pixels and/or almost no points in a lidar point cloud. Similarly, a rider on a bicycle is mostly static -- their arms aren't swinging like somebody walking, and while their legs are moving they're moving less visibly than a walking human. In addition bicycles move quickly and need to be detected at a distance more like the distance where you would need to detect a vehicle than the distance where you can reasonably detect a pedestrian.
Fortunately, mostly none of that is true when I ride a bicycle.
The kids riding skateboards and hoover boards in the traffic lane are probably going to die regardless.
Ooh, and ebikes mean that cyclists may not be moving to pedal at all.
You can get the same effect with one side of a hill.
In general if I was going to communicate one thing people should understand about autonomous vehicles, it's that the way they see the world and react to it is very, very different from how humans see the world. A vehicle is just a box with the label "vehicle", a pedestrian is (very often) just a box with the label "pedestrian". They lack any deep semantic understanding of scenes. On the other hand, they are able to react much more quickly than a human, and are able to extract 360 degree information about scene in complete darkness. They are just really different.
So getting an intuition for what "safer" and "less safe" and "good enough" means isn't as simple as using your intuition about what those things mean for a human driver. For instance, you could imagine a situation where an autonomous vehicle does not detect a pedestrian emerging from behind a vehicle until that pedestrian is only 30% occluded by the vehicle, too late to stop at a normal rate. The vehicle then executes an emergency stop let's say 500ms more rapidly than a human driver would. This causes the autonomous vehicle to get rear-ended by a human-driven vehicle. Also, the pedestrian was emerging from behind the vehicle because they were walking around to the driver's side door to get in the vehicle, an intent that would be manifestly obvious to any human observer. Who is at fault for the rear-ending? There is a legal answer, but is that the right answer?
we're all going to go off and talk about sodomy in the sand dunes now, I guess, but thank you for the enlightenment, Robot President.
237, yes, RP, your comments are as insightful as I recall them being in those older threads.
224.1: I didn't mean to imply that. What I meant to imply (and really took for granted in 208) is that AVs in SF today are where AVs in Phoenix were a few years ago. Waymo has been giving fully driverless rides in Phoenix for a year and a half and this has seemingly made very little impression on anybody. (There is a guy whose whole YouTube channel is nothing but rides in driverless Waymo cars, if you're curious how well it works.)
224.2: Literally the only social ills I am claiming AVs will address (at the margin) are (a) road safety and (b) personal vehicle ownership. It seems to me there's a techno-dystopianism at work here where we assume the entire world will be transformed at the whim of AV companies. You won't be allowed to cross the street unless you're signed into the Pedestrian app! People will be prisoners in their own car! The poor will be forced to buy AVs on credit! I honestly don't think any of that will happen. Be vigilant, sure. But we're really far away from AV companies dictating the shape of the city of tomorrow.
The biggest social ill I'd pin on AVs in the short-to-medium term (and it's a big one!) is they will discourage transition from passenger cars to mass transit. The US record on this is just incredibly shitty going all the way back. On the one hand it's hard to put too much blame on AVs for going with the flow. On the other, every little bit helps. If we were genuinely in the position where there is a mutually-exclusive choice between mass transit and AVs, obviously we should choose mass transit. But that's not the actual choice. The kind of thinking in 200 is shitty and should not go unchallenged.
224.3: You are right to be skeptical. I hope you will be pleasantly surprised.
I'm afraid assholes will set their car to circle the block all day to save $15 in parking fees
This is what my traffic engineer friend expects and fears: this and that autonomous delivery vehicles will just circulate and VMT will go up, not down.
With the share economy, your AV can off-duty as an Uber while you're at work.
241: that is literally what Tesla has claimed will happen with their full self-driving.
I'd rather park it and let the computer mine bitcoins.
I just mentioned elsewhere that as long as road use and gas is underpriced compared to the social cost of carbon, bigtech innovation is going to move in perverse directions, maximizing use of those commons.
242: And every company besides Tesla is building taxis or delivery vehicles, not private cars.
As we know, the kinds of Americans who don't want to take the train because there might be a smelly person sitting next to them are likely to be very excited to have their car picking up strangers when they aren't in it. That's the kind of lateral thinking that has given us the Hyperloop and Neuralink!
There's nothing better than crop dusting an entire train car as you find a seat.
Is anyone still trying to do personal pods? Voltron trains that travel in groups on main rail lines but split off to personally deliver you to your destination. And you don't have to sit with others.
I guess the formal term is PRT personal rapid transit probably due to some bullshit copyright rules around Voltron.
Anyway, PRT is what they called the horizontal train-elavator thing that connected the hospitals at Duke with the parking garage.
who don't want to take the train because there might be a smelly person sitting next to them
All of Musk's initiatives are about how to get as far away from other human bodies as possible. It's completely pathological.
He's over generalizing from the person he's known the longest.
From Apt. 11D, a while back:
"My brother-in-law manages the New Jersey division of a major, international architecture company. They do office buildings in Times Square and corporate headquarters out here in the burbs. He recently attended a presentation about how the firm should deal with the upcoming changes to automobiles and was like a fanatic on the topic. ...
"The firm is convinced that we'll be fully transitional into an Uber-like shared self-driving car system in the next ten to fifteen years. The first changes will happen for trucking in five years.
"And it's much more than self-driving cars. People won't own their own vehicles. They will call for a car, like an Uber service, and then chillax inside the vehicle, even taking naps, while the car takes you to work or on vacation.
"My BIL's architecture company is already designing their office parks with this future in mind. ..."
That was January 2018, so we're more than halfway to the five years mentioned in the post.
This shows the ripple effects of assumptions about self-driving cars and how they are affecting real decisions well in advance of any implementation. It's one further example of why I find the blitheness about implications of self-driving cars so off-putting.
236: "... one thing people should understand about autonomous vehicles..."
You know what people should have to understand about autonomous automobiles? Nothing. Not one whit, not a dicky-bird. Nichts. ничто. Nada.
Build the tools around people, and if you can't do that, then stop.
246: Don't forget the Fiery Celebrity Death Tunnel.
The NYT now has a user-friendly article about how lithium mining for EV batteries will fuck everything up to underscore the general point that most automobile technology doesn't scale without huge costs. Seems like people are reading it.
Build the tools around people....
And yet, when railways were cutting edge, many people believed that the human body couldn't tolerate the acceleration. But investors kept right on building them. Once, if you wanted to drive a car on the public roads you had to have a guy walk in front of you with a red flag to warn people driving horses. But people went on buying cars.. If Musk wants to throw the budget of a small country at a hobby, I'd rather he picked AVs than gratifying a handful of nutjobs who want to commit suicide by Mars lander.
In other news, yesterday I discovered where my jetpack was. It's been issued to the Royal Marines, who can now board enemy ships by flying.
This "revelation" did nothing to TSLA price because everyone knows Elon was lying.
Well, everyone involved in buying and selling significant amounts of stock. Not necessarily everyone who bought one of them.