Roadmap for autonomous vehicles
December 1, 2020 7:08 AM Subscribe
Tesla has released a re-written version of its "Full Self Drive" software. Beta testers has been posting enthusiastic reviews. The company has pulled off the trick of getting customers to pay for and test software where they still bear responsibility as drivers. Others, such as Waymo, prefer to run trials in carefully mapped areas with driverless vehicles. Autonomy promises fewer cars on the road and , less space dedicated to parking. Perhaps manufacturers will prefer to profit from their own vehicle networks rather than by selling cars to the public? Tesla talks (contentiously) about providing systems which are much safer than imperfect human drivers - but the developing technology must cross a chasm of "edge cases" to succeed. This can be painful: there have been fatalities associated with driver's employing Tesla's auto-pilot and FSD, to date. A world of Robotaxis will also end numerous transport related jobs. But can we be confident it will never happen? The technology will surely tempt governments and planners.
Amidst this chaos of this year, I had kind of forgotten to spare any thought to this topic of self-driving cars. But now, with this reminder, the first thing that occurred to me was that with so many people off the roads during the quarantine and with the still-evolving shift toward a more remote workforce*, the integration shift to fully automated driving might have gotten just a tiny, tiny bit easier. Obviously, post-COVID (speaking of fever dreams), the roads will fill back up again, but hopefully the remote-work options will remain and even continue to grow.
I am super, super ready for a reality where accepting the potential for a horrible injury or death simply to get to work or whatever stops being a thing. Or at least, it's so greatly reduced as to be legitimately dismissible. Apologies for the morning ramble response, great post!
* For applicable industries only, terms and conditions still apply.
posted by Godspeed.You!Black.Emperor.Penguin at 7:29 AM on December 1, 2020 [1 favorite]
I am super, super ready for a reality where accepting the potential for a horrible injury or death simply to get to work or whatever stops being a thing. Or at least, it's so greatly reduced as to be legitimately dismissible. Apologies for the morning ramble response, great post!
* For applicable industries only, terms and conditions still apply.
posted by Godspeed.You!Black.Emperor.Penguin at 7:29 AM on December 1, 2020 [1 favorite]
So much of the fanboy-ism of Tesla is tied up in the cult of personality around Elon, to the point where it's easy to forget how important Tesla has been to pushing the envelope for EV technology ahead of where things would be if we were just relying on other carmakers to gradually transition away from combustion engines. The CARB/ZEV regulations do help, but without a swashbuckling billionaire willing to chase the early adopters, I don't think companies like Kia, Hyundai, VW, and now even Ford and GM would be so ambitious with their electric vehicle plans. If Tesla didn't exist, I probably wouldn't have an EV in my garage that I can drive for ~250 miles before recharging, so I'm glad it exists.
On the other hand, Elon's a fabulist snake-oil salesman who's been telling flat-out lies about the immediate and medium-term potential for fully automated driving, and he deserves to be roasted for that. The tech just isn't ready, and the "move fast, break stuff" methodology works a lot better when "break stuff" means "grandma might briefly lose access to pictures of her grandson" than when it means "grandma's dead because the car she was traveling in swerved into a tractor trailer."
posted by tonycpsu at 7:42 AM on December 1, 2020 [31 favorites]
On the other hand, Elon's a fabulist snake-oil salesman who's been telling flat-out lies about the immediate and medium-term potential for fully automated driving, and he deserves to be roasted for that. The tech just isn't ready, and the "move fast, break stuff" methodology works a lot better when "break stuff" means "grandma might briefly lose access to pictures of her grandson" than when it means "grandma's dead because the car she was traveling in swerved into a tractor trailer."
posted by tonycpsu at 7:42 AM on December 1, 2020 [31 favorites]
FFS, can't we just build proper public transit and denser cities?
posted by cman at 7:44 AM on December 1, 2020 [66 favorites]
posted by cman at 7:44 AM on December 1, 2020 [66 favorites]
I don't know why we would let our elected officials bank our future on a private entrepreneur (and one who chronically over-promises). We should be demanding land-use policies that enable all kinds of people to get to work using transit, which is operated by trained and licensed employees and has the full force of the government behind it.
I think autonomous driving technologies have potential in the trucking industry, but only as an assist for a human driver. And even in this case, the presence of the technology may increase the risk of crashes if implemented too soon or if the workforce is not trained on how to use the system's assistance without being derelict on their own responsibility on the road.
In a crowded urban setting, not only is autonomous driving technology an imminent danger to pedestrians, children, the elderly, cyclists, but it enables the further congestion of our public space by vehicles carrying one person and discourages efficient planning.
Not to mention that it will effectively be a plaything for the rich, leaving average workers and families to rely on a deprecated infrastructure and aging personal vehicles which financial responsibility falls on those least able to meet it.
posted by ism at 7:46 AM on December 1, 2020 [9 favorites]
I think autonomous driving technologies have potential in the trucking industry, but only as an assist for a human driver. And even in this case, the presence of the technology may increase the risk of crashes if implemented too soon or if the workforce is not trained on how to use the system's assistance without being derelict on their own responsibility on the road.
In a crowded urban setting, not only is autonomous driving technology an imminent danger to pedestrians, children, the elderly, cyclists, but it enables the further congestion of our public space by vehicles carrying one person and discourages efficient planning.
Not to mention that it will effectively be a plaything for the rich, leaving average workers and families to rely on a deprecated infrastructure and aging personal vehicles which financial responsibility falls on those least able to meet it.
posted by ism at 7:46 AM on December 1, 2020 [9 favorites]
Autonomous vehicles are inevitable. Even if you don't stamp out all of the edge cases, your software and hardware platform only has to be better than people at driving. And people are really crappy drivers.
That said, these wild west games that companies like Tesla and Uber are playing need to be stopped dead in their tracks, and they need to be forced to conform to very reasonable automation development roadmaps places like California already have established.
It's long overdue to pierce the corporate veil for malfeasance on the part of executives, and hold them criminally accountable for their reckless behavior. Uber is throwing their driver under the bus for the Arizona death, but I don't see a single executive up on charges.
Without real and public accountability, the public will be even slower to accept autonomous vehicles.
posted by tclark at 7:48 AM on December 1, 2020 [1 favorite]
That said, these wild west games that companies like Tesla and Uber are playing need to be stopped dead in their tracks, and they need to be forced to conform to very reasonable automation development roadmaps places like California already have established.
It's long overdue to pierce the corporate veil for malfeasance on the part of executives, and hold them criminally accountable for their reckless behavior. Uber is throwing their driver under the bus for the Arizona death, but I don't see a single executive up on charges.
Without real and public accountability, the public will be even slower to accept autonomous vehicles.
posted by tclark at 7:48 AM on December 1, 2020 [1 favorite]
Robotic cars have too many videos and personal accounts for any but flat-earther level deniability that they can and do work in many places. Do they work well enough to be allowed on public roads? Statistically they seem much safer than human drivers, but will the news-cycle worthy, rare, horrific, exceptional example prevent the technology from acceptance?
Not much of a flyer but I would not have worried about boarding a 737 in the weeks after the horrific accidents unless it was on a poor/sketchy airline that skipped the essential safety features. If you read about it in the news, don't worry about it personally --- news worthy events are rare, worry about crossing the road, driving tired or impaired within 5 miles of your home. 38,800 auto deaths in 2019, the 50 or so involving a tesla included the other car driving the wrong way or other reasons not the teslas fault.
In 20 or so years a car without safety sensors will probably be forbidden on public roads. The level of technology for self driving cars in mind boggling but once a program works, copying it to all the other devices/cars is economically at marginal cost free (many non trivial hard problems but the marginal cost holds). At the moment cost of hardware is probably the main issue, automotive sensors and cameras are not just pricey but not even available at volume. Once the right devices are built at low cost/high quantity many industries will become robotized.
posted by sammyo at 7:52 AM on December 1, 2020 [1 favorite]
Not much of a flyer but I would not have worried about boarding a 737 in the weeks after the horrific accidents unless it was on a poor/sketchy airline that skipped the essential safety features. If you read about it in the news, don't worry about it personally --- news worthy events are rare, worry about crossing the road, driving tired or impaired within 5 miles of your home. 38,800 auto deaths in 2019, the 50 or so involving a tesla included the other car driving the wrong way or other reasons not the teslas fault.
In 20 or so years a car without safety sensors will probably be forbidden on public roads. The level of technology for self driving cars in mind boggling but once a program works, copying it to all the other devices/cars is economically at marginal cost free (many non trivial hard problems but the marginal cost holds). At the moment cost of hardware is probably the main issue, automotive sensors and cameras are not just pricey but not even available at volume. Once the right devices are built at low cost/high quantity many industries will become robotized.
posted by sammyo at 7:52 AM on December 1, 2020 [1 favorite]
Cars are not the solution. They take up too damn much space whether they are driverless or not. Our cities have been destroyed by making space for cars. We have had a solution in hand for 100 years (density! Trains! Busses!) but refuse to use it, basically because white people don’t want to share space with black people.
posted by rockindata at 7:52 AM on December 1, 2020 [35 favorites]
posted by rockindata at 7:52 AM on December 1, 2020 [35 favorites]
I understand that statistically, full self driving is probably already safer than having a random human behind the wheel, but I'm not a random human, I am me, so I'd rather have my fate in my own hands as much as possible. Millions of people drive by oddly painted lane markers that'll steer you into walls without dying, but if the algorithm gets it wrong you're toast.
The problem is that fully automated driving is terrible until it's perfect and it's going to take a long long time figuring out how to get between the two without massive numbers of casualties (yes, we already have massive numbers of casualties with human drivers).
That being said, I love Teslas as cars, despite their faults, but I can't see myself giving Elon any money. I'm waiting for another manufacturer to come along and put cars out at their level, or for a traditional car manufacturer to figure out why Tesla is so successful at what they do, so they'll release something that's not a weird plugin hybrid, or something that looks like an electric car, or some unobtainable hypercar thing. Give me something that looks sporty, has great range and is fun to drive for under $100k.
Also, agreed on real public transit and high speed trains and denser cities, not hyperloop or whatever boondoggles are going to personally enrich Elon and his ilk.
posted by mikesch at 7:54 AM on December 1, 2020
The problem is that fully automated driving is terrible until it's perfect and it's going to take a long long time figuring out how to get between the two without massive numbers of casualties (yes, we already have massive numbers of casualties with human drivers).
That being said, I love Teslas as cars, despite their faults, but I can't see myself giving Elon any money. I'm waiting for another manufacturer to come along and put cars out at their level, or for a traditional car manufacturer to figure out why Tesla is so successful at what they do, so they'll release something that's not a weird plugin hybrid, or something that looks like an electric car, or some unobtainable hypercar thing. Give me something that looks sporty, has great range and is fun to drive for under $100k.
Also, agreed on real public transit and high speed trains and denser cities, not hyperloop or whatever boondoggles are going to personally enrich Elon and his ilk.
posted by mikesch at 7:54 AM on December 1, 2020
Fully automated driving seems like a foregone conclusion for freeways and especially freight. I don’t think it has a viable future inside cities, where transportation planners are biasing toward pedestrian, cycling, and transit uses that all make autonomy significantly harder by putting more obstacles and distractions in the road to get cars moving slower.
posted by migurski at 7:57 AM on December 1, 2020
posted by migurski at 7:57 AM on December 1, 2020
> FFS, can't we just build proper public transit and denser cities?
How does this strategy play out in the light of the past year? Not being snarky, just genuinely curious as to how we address this problem. Ideally, with enough quality education, fighting the firehose of propaganda, etc would get us to a point where more people would adhere to the relevant safety guidelines during a pandemic, but that fever dream phrase just popped up again for some reason.
posted by Godspeed.You!Black.Emperor.Penguin at 7:59 AM on December 1, 2020 [5 favorites]
How does this strategy play out in the light of the past year? Not being snarky, just genuinely curious as to how we address this problem. Ideally, with enough quality education, fighting the firehose of propaganda, etc would get us to a point where more people would adhere to the relevant safety guidelines during a pandemic, but that fever dream phrase just popped up again for some reason.
posted by Godspeed.You!Black.Emperor.Penguin at 7:59 AM on December 1, 2020 [5 favorites]
I have a Kia Stinger which has driving assistance up to maybe level 1.9 and I find it to be tiring to drive long distances in my wife's car which has no driving assistance at all. Even if it's not FSD, it can lower the amount of fatigue that driving causes on longer trips.
Highways and freeways are just long, straight pieces of flat road so self-driving on those seems inevitable and should probably be encouraged. Meanwhile, pedestrian avoidance systems just keep getting better which should make things much safer in cities in the long run.
The biggest problem with public transport for most people is that it takes way too long to get places in the US. Back home I used to use public transport every day for work. I would walk the half mile to the station, get on my train into the city, walk to the other platform, then catch a train back out to Subiaco and walk the quarter mile to work from the station. It was about 10 minutes more than driving (45 vs 35) due to the traffic on the freeway but the ten extra minutes I could spend doing what I wanted.
Back in Northern California? Going 7 miles took me 15 minutes to drive up San Tomas Expressway. Taking a couple of buses would take the better part of an hour. There's no question which way I would take to work. Plus since my wife worked close by we would car pool and skip even more of the morning traffic.
The smart thing to do would be to use an Uber like algorithm to make dynamic public transport routes based on demand. You put your commute in, an electric bus comes to the closest arterial road which have all been marked with stop numbers, you hop on, it drops you off at the stop on the nearest arterial road to your destination. Maybe an extra 10 minutes instead of 40. That I could be convinced to use.
posted by Your Childhood Pet Rock at 8:09 AM on December 1, 2020
Highways and freeways are just long, straight pieces of flat road so self-driving on those seems inevitable and should probably be encouraged. Meanwhile, pedestrian avoidance systems just keep getting better which should make things much safer in cities in the long run.
The biggest problem with public transport for most people is that it takes way too long to get places in the US. Back home I used to use public transport every day for work. I would walk the half mile to the station, get on my train into the city, walk to the other platform, then catch a train back out to Subiaco and walk the quarter mile to work from the station. It was about 10 minutes more than driving (45 vs 35) due to the traffic on the freeway but the ten extra minutes I could spend doing what I wanted.
Back in Northern California? Going 7 miles took me 15 minutes to drive up San Tomas Expressway. Taking a couple of buses would take the better part of an hour. There's no question which way I would take to work. Plus since my wife worked close by we would car pool and skip even more of the morning traffic.
The smart thing to do would be to use an Uber like algorithm to make dynamic public transport routes based on demand. You put your commute in, an electric bus comes to the closest arterial road which have all been marked with stop numbers, you hop on, it drops you off at the stop on the nearest arterial road to your destination. Maybe an extra 10 minutes instead of 40. That I could be convinced to use.
posted by Your Childhood Pet Rock at 8:09 AM on December 1, 2020
City density has not been shown to be a deciding factor in covid spread in the US.
Maybe if we had a society where we recognized ourselves as part of a larger community, then we could have reduced covid spread in the US. Instead we see ourselves as atoms only responsible for ourselves. Promoting single-occupancy vehicles is a reflection of that mentality.
posted by tofu_crouton at 8:14 AM on December 1, 2020 [19 favorites]
Maybe if we had a society where we recognized ourselves as part of a larger community, then we could have reduced covid spread in the US. Instead we see ourselves as atoms only responsible for ourselves. Promoting single-occupancy vehicles is a reflection of that mentality.
posted by tofu_crouton at 8:14 AM on December 1, 2020 [19 favorites]
One way I thought of was when jammed on an over filled bus, why don't they stage overflow buses half way on busy routes? Mostly cost of labor. Now an optimized robot bus could be economic and effective, keeping the costs comfort of busy routes. (Here in Boston the farsighted </sarcasm> transit authority is essentially using covid to shut down a bunch of routes).
posted by sammyo at 8:16 AM on December 1, 2020
posted by sammyo at 8:16 AM on December 1, 2020
In these conversations on Metafilter, we always argue with the public transit we currently have (under an austerity administration) against the cool theoretical cars and roadways of the future.
posted by tofu_crouton at 8:17 AM on December 1, 2020 [40 favorites]
posted by tofu_crouton at 8:17 AM on December 1, 2020 [40 favorites]
In about fifty years the commute to work will look a lot more like this than this.
posted by aramaic at 8:18 AM on December 1, 2020 [3 favorites]
posted by aramaic at 8:18 AM on December 1, 2020 [3 favorites]
Fully autonomous cars have been 5 - 10 years away for at least 15 years now and I'd be shocked if they were actually working reliably by even 2030.
posted by octothorpe at 8:25 AM on December 1, 2020 [5 favorites]
posted by octothorpe at 8:25 AM on December 1, 2020 [5 favorites]
We can see the system deal with poor visibility, roundabouts and busy city streets in what looks like an effective manner.
I'd never seen those live, in-dash, wire-frame visualizations before. They look so cool. They look futuristic in the way the future was imagined when I was a kid in the 1980s. I love it.
posted by paper chromatographologist at 8:26 AM on December 1, 2020 [2 favorites]
I'd never seen those live, in-dash, wire-frame visualizations before. They look so cool. They look futuristic in the way the future was imagined when I was a kid in the 1980s. I love it.
posted by paper chromatographologist at 8:26 AM on December 1, 2020 [2 favorites]
even the most normal, boring cars (Honda Civic, Toyota Corolla) now have a higher-end package with auto-lanekeeping + adaptive cruise control. this will probably trickle down to base models, possibly due to regulation like with backup cameras.
those two features -- turning the wheel to keep you in the lane, and braking/accelerating to keep your distance to the car in front -- are sort of an 80/20 thing compared to full self driving -- they are simple and reliable but cover enough of the experience to seem pretty magical, especially on highways. you can't stop paying attention but it's surprising how much they reduce fatigue.
this + automatic lane changes on highways seems like it will probably become much more common in the near future -- hard to see downside.
Tesla Beta FSD... I don't know. I would have said people wouldn't put up with it before after a few spectacular accidents, but one thing we've learned this year is many Americans will tolerate a lot of death for the sake of minor lifestyle conveniences... and the argument that it net prevents accidents may still be credible.
posted by vogon_poet at 8:29 AM on December 1, 2020 [1 favorite]
those two features -- turning the wheel to keep you in the lane, and braking/accelerating to keep your distance to the car in front -- are sort of an 80/20 thing compared to full self driving -- they are simple and reliable but cover enough of the experience to seem pretty magical, especially on highways. you can't stop paying attention but it's surprising how much they reduce fatigue.
this + automatic lane changes on highways seems like it will probably become much more common in the near future -- hard to see downside.
Tesla Beta FSD... I don't know. I would have said people wouldn't put up with it before after a few spectacular accidents, but one thing we've learned this year is many Americans will tolerate a lot of death for the sake of minor lifestyle conveniences... and the argument that it net prevents accidents may still be credible.
posted by vogon_poet at 8:29 AM on December 1, 2020 [1 favorite]
how safe can we make autonomous cars? Likely, with time and diligence, very very very very very safe. But I suspect if we let the free market do its thing, what will actually happen is that we'll end up making them as unsafe as we can legally get away with. Do we even have to make them as safe as the old school automobile? Or just the ones that rich people and their loved ones will be using?
count me in the "let's try to imagine a future that doesn't require a ubiquity of cars, self-driving or otherwise" crowd. I personally managed to live over twenty years without owning one myself between 1993-2016, a situation that only changed when I was forced to re-locate to a rural location. Cars are like horses were back in the early days of the 20th century -- no longer required for our cities to function, a hindrance in fact. Let's keep them in the country.
posted by philip-random at 8:38 AM on December 1, 2020 [2 favorites]
count me in the "let's try to imagine a future that doesn't require a ubiquity of cars, self-driving or otherwise" crowd. I personally managed to live over twenty years without owning one myself between 1993-2016, a situation that only changed when I was forced to re-locate to a rural location. Cars are like horses were back in the early days of the 20th century -- no longer required for our cities to function, a hindrance in fact. Let's keep them in the country.
posted by philip-random at 8:38 AM on December 1, 2020 [2 favorites]
Your Childhood Pet Rock: Meanwhile, pedestrian avoidance systems just keep getting better which should make things much safer in cities in the long run.
I find this interesting because every visualization of the self-driving car future treats pedestrians as an afterthought if it even thinks about them at all. I have never seen any pro-self-driving car presentation, video, explainer, pamphlet, or candy-bar wrapper that factors in pedestrians as a valid mode of transportation. (Along with bikes, mass transit, etc.) It's absurd. I mean, something as simple as "I want to cross the street to go to the store," which assumes you live somewhere that has a store across the street, is absolutely not factored in. It's just "look at how smooth all this traffic flows now!" There's no consideration about pedestrians crossing the street. None.
I've had this argument with self-driving car wonks and they love to propose stuff like "Oh, just build pedestrian bridges," as if that isn't an accessibility nightmare. Pedestrian bridges are hot garbage for people in wheelchairs, on crutches, pushing strollers or carts, or just have trouble walking up and down stairs and ramps but are otherwise physically able to navigate the world.
Self-driving cars solve the wrong problem at the wrong level. Safety is one thing, and I'm also not convinced we'll have self-driving cars that can manage a winter nor'easter or heavy summer thunderstorm any time soon—the sorts of things that challenge even really good drivers.
We. Need. Mass Transit. We. Need. Multi-Modal. Transit. Not just self-driving cars. Not Elon's Boring Company Loop shit. Self-driving cars are not a solution, they are a hack that hides the real problems of individualized transit.
posted by SansPoint at 8:43 AM on December 1, 2020 [19 favorites]
I find this interesting because every visualization of the self-driving car future treats pedestrians as an afterthought if it even thinks about them at all. I have never seen any pro-self-driving car presentation, video, explainer, pamphlet, or candy-bar wrapper that factors in pedestrians as a valid mode of transportation. (Along with bikes, mass transit, etc.) It's absurd. I mean, something as simple as "I want to cross the street to go to the store," which assumes you live somewhere that has a store across the street, is absolutely not factored in. It's just "look at how smooth all this traffic flows now!" There's no consideration about pedestrians crossing the street. None.
I've had this argument with self-driving car wonks and they love to propose stuff like "Oh, just build pedestrian bridges," as if that isn't an accessibility nightmare. Pedestrian bridges are hot garbage for people in wheelchairs, on crutches, pushing strollers or carts, or just have trouble walking up and down stairs and ramps but are otherwise physically able to navigate the world.
Self-driving cars solve the wrong problem at the wrong level. Safety is one thing, and I'm also not convinced we'll have self-driving cars that can manage a winter nor'easter or heavy summer thunderstorm any time soon—the sorts of things that challenge even really good drivers.
We. Need. Mass Transit. We. Need. Multi-Modal. Transit. Not just self-driving cars. Not Elon's Boring Company Loop shit. Self-driving cars are not a solution, they are a hack that hides the real problems of individualized transit.
posted by SansPoint at 8:43 AM on December 1, 2020 [19 favorites]
The space shuttle had four computers all acting in parallel. All four results were compared and if one of the computers was off, it was ignored. Sort of a computational majority rules.
My main concern about automatic driving is the technology. I’ve been around computers since the late 70’s and, yes, they are a whole lot more powerful these days. But overall reliability seems to have gone down. Add in the extremely large and complex software and the reliability decreases even more.
So how much redundancy will be built in to these auto driving systems? How much error checking? What happens if any required component in the system fails. Hey, maybe the blue screen of death will no longer be a metaphor.
posted by njohnson23 at 8:47 AM on December 1, 2020
My main concern about automatic driving is the technology. I’ve been around computers since the late 70’s and, yes, they are a whole lot more powerful these days. But overall reliability seems to have gone down. Add in the extremely large and complex software and the reliability decreases even more.
So how much redundancy will be built in to these auto driving systems? How much error checking? What happens if any required component in the system fails. Hey, maybe the blue screen of death will no longer be a metaphor.
posted by njohnson23 at 8:47 AM on December 1, 2020
I find this interesting because every visualization of the self-driving car future treats pedestrians as an afterthought if it even thinks about them at all. I have never seen any pro-self-driving car presentation, video, explainer, pamphlet, or candy-bar wrapper that factors in pedestrians as a valid mode of transportation.
The IIHS, Euro NCAP, and ANCAP all test automated emergency braking safety systems professionally. Some models perform better than others but god damn once they get them right they are way better than most humans are at the task.
posted by Your Childhood Pet Rock at 8:51 AM on December 1, 2020 [1 favorite]
The IIHS, Euro NCAP, and ANCAP all test automated emergency braking safety systems professionally. Some models perform better than others but god damn once they get them right they are way better than most humans are at the task.
posted by Your Childhood Pet Rock at 8:51 AM on December 1, 2020 [1 favorite]
Just because Elon isn't going out and extolling the virtues doesn't mean there aren't tens of thousands of engineers toiling away in the background using technology to make our roads much safer. These are the people who are actually doing the incremental improvement slog of automated driving systems.
posted by Your Childhood Pet Rock at 8:52 AM on December 1, 2020 [1 favorite]
posted by Your Childhood Pet Rock at 8:52 AM on December 1, 2020 [1 favorite]
>can't we just build proper public transit and denser cities?
the only reason I put $100 down on a trimotor preorder is to get something that will move me and ~2000lbs of my stuff as far from my fellow fuckos as humanly possible (and my $/PTO budgets allow).
Tesla's FSD effort ties into this nicely; later this decade I might be able to:
Monday: work at office 7AM - 6PM
Tuesday: work at office 7AM - 6PM
Wednesday: work at office 7AM - 11AM, relocate ~500+ miles away that PM (total cost: $0)
Thursday: work at remote location 7AM - 6PM (hopefully w/ Starlink internet)
Friday: PTO at remote location / elsewhere
Satuday: drive home in the PM
Sunday: relax
posted by Heywood Mogroot III at 8:55 AM on December 1, 2020
the only reason I put $100 down on a trimotor preorder is to get something that will move me and ~2000lbs of my stuff as far from my fellow fuckos as humanly possible (and my $/PTO budgets allow).
Tesla's FSD effort ties into this nicely; later this decade I might be able to:
Monday: work at office 7AM - 6PM
Tuesday: work at office 7AM - 6PM
Wednesday: work at office 7AM - 11AM, relocate ~500+ miles away that PM (total cost: $0)
Thursday: work at remote location 7AM - 6PM (hopefully w/ Starlink internet)
Friday: PTO at remote location / elsewhere
Satuday: drive home in the PM
Sunday: relax
posted by Heywood Mogroot III at 8:55 AM on December 1, 2020
How does this strategy play out in the light of the past year? Not being snarky, just genuinely curious as to how we address this problem.
By this past year, you mean the pandemic that has infected 1.7% of the people in San Francisco and 10.0% of the people in North Dakota? Not sure urban density is all that much of a problem.
posted by Superilla at 8:56 AM on December 1, 2020 [10 favorites]
By this past year, you mean the pandemic that has infected 1.7% of the people in San Francisco and 10.0% of the people in North Dakota? Not sure urban density is all that much of a problem.
posted by Superilla at 8:56 AM on December 1, 2020 [10 favorites]
Your Childhood Pet Rock: Pedestrian safety is about way, way more than braking. Being able to stop on a dime to save the life of someone walking or on a bike is the minimum we should expect from self-driving cars. What truly improves pedestrian and bicyclist safety on the roads is redesigning roads to de-prioritize cars, and none of the infrastructural changes around self-driving cars do that. It's about treating pedestrians as equal and valid users of the road, and part of that is deprioritizing drivers. Self-driving cars elevate the car above all other forms of transportation, if not by intent, then certainly through neglect of other modes of transportation.
posted by SansPoint at 9:19 AM on December 1, 2020 [17 favorites]
posted by SansPoint at 9:19 AM on December 1, 2020 [17 favorites]
Wow. You just hauled those goal posts right on out of the stadium.
posted by Your Childhood Pet Rock at 9:26 AM on December 1, 2020 [4 favorites]
posted by Your Childhood Pet Rock at 9:26 AM on December 1, 2020 [4 favorites]
I understand that statistically, full self driving is probably already safer than having a random human behind the wheel, but I'm not a random human, I am me, so I'd rather have my fate in my own hands as much as possible.
The problem is that when you drive, you're not just taking your own fate into your own hands -- it's also the fate of the other drivers, passengers, bicyclists and pedestrians around you.
Obviously, self-driving cars are currently a lot worse than human drivers at understanding the world around them and making good judgments. But they have huge inherent safety advantages when it comes to attentiveness and reaction time. At some point in the future, as the technology improves, the balance will probably shift to the point where autonomous vehicles are unequivocally less likely to kill or injure bystanders than 90%, or 99%, of drivers. And at that point, "self-determination" will stop being a good excuse to expose other people to unnecessary danger.
By analogy: it's totally reasonable to not want to take a COVID-19 vaccine before it's been tested for safety and effectiveness. But once it passes those tests, going out in public without being vaccinated (assuming it's reasonably possible for you to do so) starts to look like a socially irresponsible choice, even if you're not personally concerned about your own health.
I find this interesting because every visualization of the self-driving car future treats pedestrians as an afterthought if it even thinks about them at all. [...] I mean, something as simple as "I want to cross the street to go to the store," which assumes you live somewhere that has a store across the street, is absolutely not factored in. It's just "look at how smooth all this traffic flows now!" There's no consideration about pedestrians crossing the street. None.
I don't really get your point -- are you referring to stuff like that "autonomous intersection" research project that's been bouncing around for a while? There are many issues with that idea, but it seems like an entirely separate problem from the development of self-driving cars themselves. Self-driving cars could enable fancy, high-speed, pedestrian-hostile intersections, but they don't require them. Even if 100% of the cars on the road were autonomous, they could still be programmed (and legally required) to obey traffic lights, stop at crosswalks, and so on. If anything, it seems like the behavior of software is at least theoretically amenable to regulation, in a way that the behavior of humans isn't.
I don't buy that working on self-driving cars automatically implies "neglect" of other transportation options.
posted by teraflop at 9:27 AM on December 1, 2020 [1 favorite]
The problem is that when you drive, you're not just taking your own fate into your own hands -- it's also the fate of the other drivers, passengers, bicyclists and pedestrians around you.
Obviously, self-driving cars are currently a lot worse than human drivers at understanding the world around them and making good judgments. But they have huge inherent safety advantages when it comes to attentiveness and reaction time. At some point in the future, as the technology improves, the balance will probably shift to the point where autonomous vehicles are unequivocally less likely to kill or injure bystanders than 90%, or 99%, of drivers. And at that point, "self-determination" will stop being a good excuse to expose other people to unnecessary danger.
By analogy: it's totally reasonable to not want to take a COVID-19 vaccine before it's been tested for safety and effectiveness. But once it passes those tests, going out in public without being vaccinated (assuming it's reasonably possible for you to do so) starts to look like a socially irresponsible choice, even if you're not personally concerned about your own health.
I find this interesting because every visualization of the self-driving car future treats pedestrians as an afterthought if it even thinks about them at all. [...] I mean, something as simple as "I want to cross the street to go to the store," which assumes you live somewhere that has a store across the street, is absolutely not factored in. It's just "look at how smooth all this traffic flows now!" There's no consideration about pedestrians crossing the street. None.
I don't really get your point -- are you referring to stuff like that "autonomous intersection" research project that's been bouncing around for a while? There are many issues with that idea, but it seems like an entirely separate problem from the development of self-driving cars themselves. Self-driving cars could enable fancy, high-speed, pedestrian-hostile intersections, but they don't require them. Even if 100% of the cars on the road were autonomous, they could still be programmed (and legally required) to obey traffic lights, stop at crosswalks, and so on. If anything, it seems like the behavior of software is at least theoretically amenable to regulation, in a way that the behavior of humans isn't.
I don't buy that working on self-driving cars automatically implies "neglect" of other transportation options.
posted by teraflop at 9:27 AM on December 1, 2020 [1 favorite]
Actually the first presentation (by the then head of googles program) emphasized the safety, gave two life examples, a car handling a woman in a wheel chair chasing ducks in the middle of the road and a cyclist at night in a busy left turn situation where the presenter said he probably would have hit the (clearly irresponsible) cyclist.
posted by sammyo at 9:28 AM on December 1, 2020 [1 favorite]
posted by sammyo at 9:28 AM on December 1, 2020 [1 favorite]
As always, fuck this noise. Give us Trains. Trains Trains Trains Trains Trains Trains Trains.
posted by aspersioncast at 9:48 AM on December 1, 2020 [21 favorites]
posted by aspersioncast at 9:48 AM on December 1, 2020 [21 favorites]
→
Not really. The car industry has a habit of going all Calvinball with the regulations when it suits them. Early cars had a problem with killing pedestrians. Solution? Make "jaywalking" the problem! Before cars, the roads were for people. Musk has a shit-tonne of money and will magically make self-driving cars killing people someone else's problem.
And it's always been like this. The first person killed by a train, William Huskisson in 1830 (on the first day that trains ever ran), has been variously described as infirm and/or deaf: standard victim blaming stuff. On the other hand, I've never heard of Stephenson's Rocket being described as "unfit for purpose" or having "inadequate brakes". Like the Musk of his day, Stephenson had his fanbois, always eager to scribble that it was people like Huskisson who got in the way of progress, not that the new tech was problematic in any way.
posted by scruss at 10:00 AM on December 1, 2020 [14 favorites]
Wow. You just hauled those goal posts right on out of the stadium.
Not really. The car industry has a habit of going all Calvinball with the regulations when it suits them. Early cars had a problem with killing pedestrians. Solution? Make "jaywalking" the problem! Before cars, the roads were for people. Musk has a shit-tonne of money and will magically make self-driving cars killing people someone else's problem.
And it's always been like this. The first person killed by a train, William Huskisson in 1830 (on the first day that trains ever ran), has been variously described as infirm and/or deaf: standard victim blaming stuff. On the other hand, I've never heard of Stephenson's Rocket being described as "unfit for purpose" or having "inadequate brakes". Like the Musk of his day, Stephenson had his fanbois, always eager to scribble that it was people like Huskisson who got in the way of progress, not that the new tech was problematic in any way.
posted by scruss at 10:00 AM on December 1, 2020 [14 favorites]
So much of the fanboy-ism of Tesla is tied up in the cult of personality around Elon, to the point where it's easy to forget how important Tesla has been to pushing the envelope for EV technology ahead of where things would be if we were just relying on other carmakers to gradually transition away from combustion engines.
My father has three electric cars in his driveway, two of which were sold before Tesla existed as a company, the other is a 2019 model both faster and more luxurious than the comparable Tesla model, but built by a company that sells more cars by the third week of January than Team Elon sells all year.
Tesla has pushed the envelope on people never shutting the fuck up about Tesla, and that's about it really. Now the big boys are coming into the market, and (God willing) we will soon see Tesla driven into the dust and we'll be forced to hear from Elon ever, ever, again.
posted by sideshow at 10:01 AM on December 1, 2020 [8 favorites]
My father has three electric cars in his driveway, two of which were sold before Tesla existed as a company, the other is a 2019 model both faster and more luxurious than the comparable Tesla model, but built by a company that sells more cars by the third week of January than Team Elon sells all year.
Tesla has pushed the envelope on people never shutting the fuck up about Tesla, and that's about it really. Now the big boys are coming into the market, and (God willing) we will soon see Tesla driven into the dust and we'll be forced to hear from Elon ever, ever, again.
posted by sideshow at 10:01 AM on December 1, 2020 [8 favorites]
> My father has three electric cars in his driveway, two of which were sold before Tesla existed as a company, the other is a 2019 model both faster and more luxurious than the comparable Tesla model, but built by a company that sells more cars by the third week of January than Team Elon sells all year.
Tesla has pushed the envelope on people never shutting the fuck up about Tesla, and that's about it really. Now the big boys are coming into the market, and (God willing) we will soon see Tesla driven into the dust and we'll be forced to hear from Elon ever, ever, again.
"Faster" and "more luxurious" aren't the metrics on which I'm evaluating the company's impact on the market. They've been leaders on range and availability of charging stations, both of which matter a lot more to mass sales (and mass emissions reduction) than a sub 5-second 0-60 time or the quality of the leather seats. The fact that the charging stations are only for Tesla owners is a shitty vendor lock-in tactic, but it's worked to make finding a Tesla station a lot easier than it is for me to find an EVGO, ChargePoint, or Blink station when my Kia Niro EV is low on electrons.
My point is that by targeting the higher end of the early adopters and making their cars much better on a lot of metrics than the earlier generation Volts and Leafs and what-nots, Tesla gave the manufacturers more incentive to build the Kona EVs, Niro EVs, VW ID.3s, and so on, none of which can currently touch Tesla's lower-end models on many of these metrics. I respect that, even if I generally dislike the company and especially the founder.
posted by tonycpsu at 10:11 AM on December 1, 2020
Tesla has pushed the envelope on people never shutting the fuck up about Tesla, and that's about it really. Now the big boys are coming into the market, and (God willing) we will soon see Tesla driven into the dust and we'll be forced to hear from Elon ever, ever, again.
"Faster" and "more luxurious" aren't the metrics on which I'm evaluating the company's impact on the market. They've been leaders on range and availability of charging stations, both of which matter a lot more to mass sales (and mass emissions reduction) than a sub 5-second 0-60 time or the quality of the leather seats. The fact that the charging stations are only for Tesla owners is a shitty vendor lock-in tactic, but it's worked to make finding a Tesla station a lot easier than it is for me to find an EVGO, ChargePoint, or Blink station when my Kia Niro EV is low on electrons.
My point is that by targeting the higher end of the early adopters and making their cars much better on a lot of metrics than the earlier generation Volts and Leafs and what-nots, Tesla gave the manufacturers more incentive to build the Kona EVs, Niro EVs, VW ID.3s, and so on, none of which can currently touch Tesla's lower-end models on many of these metrics. I respect that, even if I generally dislike the company and especially the founder.
posted by tonycpsu at 10:11 AM on December 1, 2020
Wow. You just hauled those goal posts right on out of the stadium.
in the context of the supremacy of the automobile/autonomous-mobile continuing to define the future of our urban (and other) planning (and thinking) perhaps more than any other factor, I say good, and keep on moving them way the f*** out of the stadium, all the way across the county line if necessary.
posted by philip-random at 10:17 AM on December 1, 2020 [11 favorites]
in the context of the supremacy of the automobile/autonomous-mobile continuing to define the future of our urban (and other) planning (and thinking) perhaps more than any other factor, I say good, and keep on moving them way the f*** out of the stadium, all the way across the county line if necessary.
posted by philip-random at 10:17 AM on December 1, 2020 [11 favorites]
Team "fuck this noise" here. Give us buses, trains, electric bicycles and walkable, dense cities. Society has progressed past the need for cars.
posted by Tom-B at 10:19 AM on December 1, 2020 [10 favorites]
posted by Tom-B at 10:19 AM on December 1, 2020 [10 favorites]
As someone who's been in software for a long time, all I can say is: I'm never getting into one of those things. No way, no how. This is just how software is. It's inexact by nature. For those cases where you really need it to not kill people, you have to introduce triple-redundancy, and introduce Formal Methods just to make sure that the code you're writing does what it's supposed to in the first place. And, that only removes the obvious problems. All that extra stuff is time-consuming and expensive. And, There isn't much of a movement to do what needs to be done to even make this area remotely safe.
No way. No how.
posted by Citrus at 10:29 AM on December 1, 2020 [7 favorites]
No way. No how.
posted by Citrus at 10:29 AM on December 1, 2020 [7 favorites]
> Team "fuck this noise" here. Give us buses, trains, electric bicycles and walkable, dense cities. Society has progressed past the need for cars.
I want more trains and fewer cars, but we don't have time to wait for suburban and rural Americans to come around on the need to move on from car culture. I therefore believe in EVs as a bridge to carbon reduction that we can institute even as we push for more funding for better public transit and more pedestrian-friendly development plans. We can do both.
posted by tonycpsu at 10:31 AM on December 1, 2020 [5 favorites]
I want more trains and fewer cars, but we don't have time to wait for suburban and rural Americans to come around on the need to move on from car culture. I therefore believe in EVs as a bridge to carbon reduction that we can institute even as we push for more funding for better public transit and more pedestrian-friendly development plans. We can do both.
posted by tonycpsu at 10:31 AM on December 1, 2020 [5 favorites]
I love how the technology solution for "humans are bad at x" is to have humans write computer software to do x that codes all the flaws, mistakes, and bad assumptions humans make into a system that makes those same mistakes, but faster and much harder to audit.
posted by SansPoint at 10:32 AM on December 1, 2020 [18 favorites]
posted by SansPoint at 10:32 AM on December 1, 2020 [18 favorites]
> Now the big boys are coming into the market, and (God willing) we will soon see Tesla driven into the dust and we'll be forced to hear from Elon ever, ever, again.
Are you saying this because you hate Musk specifically or bro-style billionares generally? Because if it's the latter then taking Musk down a notch only helps the dozens of executives at the other auto companies who if they could would be living Musk's dream life of alpha-male billionaire (sex with rockstars, California mansions, popular guruhood). The auto industry has been steeping in and perpetuating toxic bro culture bullshit for longer than there've been computers, and Musk is changing that far less than you imagine.
posted by at by at 10:44 AM on December 1, 2020
Are you saying this because you hate Musk specifically or bro-style billionares generally? Because if it's the latter then taking Musk down a notch only helps the dozens of executives at the other auto companies who if they could would be living Musk's dream life of alpha-male billionaire (sex with rockstars, California mansions, popular guruhood). The auto industry has been steeping in and perpetuating toxic bro culture bullshit for longer than there've been computers, and Musk is changing that far less than you imagine.
posted by at by at 10:44 AM on December 1, 2020
He's certainly made it harder to avoid online, though.
posted by sagc at 10:45 AM on December 1, 2020 [1 favorite]
posted by sagc at 10:45 AM on December 1, 2020 [1 favorite]
And, more generally, he's managed to recruit people who really ought not to be automotive advocates - tech workers who commute to a campus, or work from home, or otherwise... live in a city with public transit - based on aesthetics, flamethrowers, and generally making up whatever statistics he wants to support his projects.
posted by sagc at 10:48 AM on December 1, 2020 [1 favorite]
posted by sagc at 10:48 AM on December 1, 2020 [1 favorite]
Stephenson had his fanbois
They are indeed the worst.
My ancestor's train Perseverance, which was entered in the same competition that Rocket won, was very safe indeed. It never got out of the shed on competition day.
posted by thatwhichfalls at 10:49 AM on December 1, 2020 [4 favorites]
They are indeed the worst.
My ancestor's train Perseverance, which was entered in the same competition that Rocket won, was very safe indeed. It never got out of the shed on competition day.
posted by thatwhichfalls at 10:49 AM on December 1, 2020 [4 favorites]
So how much redundancy will be built in to these auto driving systems? How much error checking? What happens if any required component in the system fails. Hey, maybe the blue screen of death will no longer be a metaphor.
Tesla's FSD circuit board uses 2 independently powered chips which feed data to the part of the logic that makes the decision about what the car should do. If the chips don't agree on the content of a frame then that frame is thrown out. So yes - there is the same voting idea that there was on the space shuttle. I'm not sure how this will have changed with the new re-write. See this explanation.
posted by rongorongo at 10:52 AM on December 1, 2020
Tesla's FSD circuit board uses 2 independently powered chips which feed data to the part of the logic that makes the decision about what the car should do. If the chips don't agree on the content of a frame then that frame is thrown out. So yes - there is the same voting idea that there was on the space shuttle. I'm not sure how this will have changed with the new re-write. See this explanation.
posted by rongorongo at 10:52 AM on December 1, 2020
here's what i suggest is the right measure of success: the car is ready on the day a fed law is passed saying the c-suite and senior engineering are liable for vehicular homicide and manslaughter if software fails - just like i am. when the c-suite and senior management agree to that, I'll bite. otherwise it's just opening day at jurassic park and all the tesla techies are crossing fingers and holding their breath.
it's not the case for software, because??? when it is the case for PEs signing off on skyscrapers, dams, and bridges; personal and professional liability is on the line.
posted by j_curiouser at 10:59 AM on December 1, 2020 [11 favorites]
it's not the case for software, because??? when it is the case for PEs signing off on skyscrapers, dams, and bridges; personal and professional liability is on the line.
posted by j_curiouser at 10:59 AM on December 1, 2020 [11 favorites]
it's not the case for software, because??? when it is the case for PEs signing off on skyscrapers, dams, and bridges; personal and professional liability is on the line.
Don't PE's currently sign off on roads? Their bad designs kill people every day. C-Suite of large-enough companies are already supposedly able to be jailed due to Sarbanes-Oxley rules. Never happens.
posted by The_Vegetables at 11:05 AM on December 1, 2020
Musk has a shit-tonne of money and will magically make self-driving cars killing people someone else's problem.
word.
posted by j_curiouser at 11:06 AM on December 1, 2020 [3 favorites]
word.
posted by j_curiouser at 11:06 AM on December 1, 2020 [3 favorites]
My commute to work is:
11 minutes by car
or 42 minutes by public transit (which includes 29 minutes of walking)
or about 4 seconds by signing on to the company VPN and starting Remote Desktop.
My spouse's commute is 17 minutes by car... and Google can't even plot a public transit route to it at all. I don't think our bus system crosses county lines. And she can't work from home.
To the grocery store, it's 13 minutes by car or 1 hour, 52 minutes each way by public transit. Granted, we could switch grocery stores and then it would "only" be 37 minutes (carrying a week's worth of groceries).
But if I knew I could keep working from home indefinitely after the pandemic relents, my spouse and I wouldn't need two cars.
Frankly, improved car efficiency -- and encouraging working from home whenever possible, nevermind what that Deutsche bank douchebag says -- looks both far more likely and more palatable in the short-medium term than completely replacing cars with better public transit.
posted by Foosnark at 11:17 AM on December 1, 2020 [3 favorites]
11 minutes by car
or 42 minutes by public transit (which includes 29 minutes of walking)
or about 4 seconds by signing on to the company VPN and starting Remote Desktop.
My spouse's commute is 17 minutes by car... and Google can't even plot a public transit route to it at all. I don't think our bus system crosses county lines. And she can't work from home.
To the grocery store, it's 13 minutes by car or 1 hour, 52 minutes each way by public transit. Granted, we could switch grocery stores and then it would "only" be 37 minutes (carrying a week's worth of groceries).
But if I knew I could keep working from home indefinitely after the pandemic relents, my spouse and I wouldn't need two cars.
Frankly, improved car efficiency -- and encouraging working from home whenever possible, nevermind what that Deutsche bank douchebag says -- looks both far more likely and more palatable in the short-medium term than completely replacing cars with better public transit.
posted by Foosnark at 11:17 AM on December 1, 2020 [3 favorites]
While I agree that these kinds of vehicles will most likely lead to fewer accidents and deaths, though I can sense the lawyers salivating at the thought of finding blame when an accident does happen. When a car does run over a dog, child, or doesn't know how to choose the best of a bad situation, there will be plenty of finger-pointing. I hope I'm wrong.
posted by cparkins at 11:33 AM on December 1, 2020
posted by cparkins at 11:33 AM on December 1, 2020
I still think autonomous vehicles are a key part in solving transit. So far our attempts at density to improve transit seem to only make housing prices get really really expensive, this only pushes those with less money to spend on housing further away from the dense core where transit options are less interesting and makes their situation even worse as they have to allocate so much time to transit. Mass transit trunks are necessary to move a lot of people quickly but there are still transport needs outside of the trunk because, once again, housing near the trunks gets really expensive.
We’re not great at solving the sparse housing to trunk problem. Its either bus rides that take forever or gigantic parkings around stations. Buses move a lot of people when full (not a given in some neighbourhoods), but need a driver, a schedule that can be crippling and even though they get you there, they take so much time since they’re always stopping to let people in/out. I still think a fleet of autonomous mini buses/vans that can schedule itself dynamically and enforce a form a pooling could get a lot of people to the mass transit trunks efficiently and in a very reasonable time.
If we truly want to reduce our CO2 emissions we’ll need solutions that don’t rely on people doing “the right thing” because I think at this point its pretty clear there’s a huge chunk of people who won’t do that, heck they won’t protect themselves from a virus that is here NOW. Climate change is an issue where consequences are very unconnected from actions, that’s even worse. If public/mass transit becomes cheaper/just as comfortable/nice as a solo car there’s a chance it’ll be adopted.
We need to solve this problem in the current reality, wishing we had different dense/walkable city layouts is not gonna make them appear out of thin air and make then affordable. And yes, I realize I'm kinda advocating for a tech that isn't there (yet?) and transit trunks that aren't always there. Although the dynamic routing is doable without autonomous cars we just need to throw some money at the problem and pay drivers.
posted by WaterAndPixels at 11:56 AM on December 1, 2020 [6 favorites]
We’re not great at solving the sparse housing to trunk problem. Its either bus rides that take forever or gigantic parkings around stations. Buses move a lot of people when full (not a given in some neighbourhoods), but need a driver, a schedule that can be crippling and even though they get you there, they take so much time since they’re always stopping to let people in/out. I still think a fleet of autonomous mini buses/vans that can schedule itself dynamically and enforce a form a pooling could get a lot of people to the mass transit trunks efficiently and in a very reasonable time.
If we truly want to reduce our CO2 emissions we’ll need solutions that don’t rely on people doing “the right thing” because I think at this point its pretty clear there’s a huge chunk of people who won’t do that, heck they won’t protect themselves from a virus that is here NOW. Climate change is an issue where consequences are very unconnected from actions, that’s even worse. If public/mass transit becomes cheaper/just as comfortable/nice as a solo car there’s a chance it’ll be adopted.
We need to solve this problem in the current reality, wishing we had different dense/walkable city layouts is not gonna make them appear out of thin air and make then affordable. And yes, I realize I'm kinda advocating for a tech that isn't there (yet?) and transit trunks that aren't always there. Although the dynamic routing is doable without autonomous cars we just need to throw some money at the problem and pay drivers.
posted by WaterAndPixels at 11:56 AM on December 1, 2020 [6 favorites]
Also that whole autonomous car systems/safety we need to treat it as we treat aviation. Regulate it heavily it's Tesla's responsibility to prove its system is safe and to pay to certify it.
posted by WaterAndPixels at 12:06 PM on December 1, 2020 [3 favorites]
posted by WaterAndPixels at 12:06 PM on December 1, 2020 [3 favorites]
One thing almost all who drive have to deal with is the hard choice to give it up when reaching a certain age. Autonomous vehicles will be a boon to seniors who are no longer able to drive safely, but want to maintain their independence.
posted by ShooBoo at 12:38 PM on December 1, 2020 [3 favorites]
posted by ShooBoo at 12:38 PM on December 1, 2020 [3 favorites]
Matchbox is on board with autonomous vehicles; Hot Wheels has invested in Tesla (Roadsters, Model S) and full solar power.
posted by JDC8 at 12:48 PM on December 1, 2020 [1 favorite]
posted by JDC8 at 12:48 PM on December 1, 2020 [1 favorite]
WaterAndPixels: Time and time again, it's been shown that "if you build it, they will ride" for Public Transit... with two caveats. Induced demand isn't just a factor that affects road capacity. It also works for transit.
1. It has to go where from where people are to where people want to go.
2. It has to run on a reliable, regular schedule.
The former is usually pretty obvious and easy to do, but you'd be surprised how often you wind up with a transit line to/from nowhere because of some schmuck politician who wants pork. The latter is the real problem. All-too-often, because of fears of low ridership and losing money (oh no, not losing money!) political pressure forces transit planners to run minimal service that doesn't serve the people who need it or would want to use it. I'm talking two-car trains, or headways of 30 minutes or more at peak hours.
Speaking of induced demand, that's another problem with self-driving cars that never gets brought up. Roads have capacity limits, and self-driving cars no matter how efficiently they manage traffic, will still create traffic that will still clog roads if there's too much volume for that road to contain. If you expand the road, induced demand will just fill that capacity in time. If you want to reduce traffic, you reduce capacity while providing alternatives to that route through public transit and other alternatives to individual cars.
posted by SansPoint at 1:19 PM on December 1, 2020 [8 favorites]
1. It has to go where from where people are to where people want to go.
2. It has to run on a reliable, regular schedule.
The former is usually pretty obvious and easy to do, but you'd be surprised how often you wind up with a transit line to/from nowhere because of some schmuck politician who wants pork. The latter is the real problem. All-too-often, because of fears of low ridership and losing money (oh no, not losing money!) political pressure forces transit planners to run minimal service that doesn't serve the people who need it or would want to use it. I'm talking two-car trains, or headways of 30 minutes or more at peak hours.
Speaking of induced demand, that's another problem with self-driving cars that never gets brought up. Roads have capacity limits, and self-driving cars no matter how efficiently they manage traffic, will still create traffic that will still clog roads if there's too much volume for that road to contain. If you expand the road, induced demand will just fill that capacity in time. If you want to reduce traffic, you reduce capacity while providing alternatives to that route through public transit and other alternatives to individual cars.
posted by SansPoint at 1:19 PM on December 1, 2020 [8 favorites]
SansPoint, I agree on induced demand and I don’t think we should be adding more roads/lanes, unless they reserved for mass/public transit. I just think self driving vans/mini buses could help if they stay local and drive you to other transit or a local destination. Especially in low density areas where other options are too expensive/difficult. It’s about providing solutions for already built housing.
There’s great potential for abuse where somebody will decide to live crazy far way and just sleep/work the transit in its self driving car and not care about traffic. And wed probably have to ban “empty cars” so that we can continue to use parking availability as a way to reduce solo car usage if self driving cars become a thing, it would probably double the traffic also if cars are driving back from dense areas to park.
posted by WaterAndPixels at 1:44 PM on December 1, 2020
There’s great potential for abuse where somebody will decide to live crazy far way and just sleep/work the transit in its self driving car and not care about traffic. And wed probably have to ban “empty cars” so that we can continue to use parking availability as a way to reduce solo car usage if self driving cars become a thing, it would probably double the traffic also if cars are driving back from dense areas to park.
posted by WaterAndPixels at 1:44 PM on December 1, 2020
There have been a couple of interesting experiments on dynamically-dispatched buses, where the bus route depends on who's onboard, and the schedule depends on how many people are asking for buses. I don't think they've gone system-wide anywhere, but it's an intriguing idea. Sort of like "Uber Pool" except for buses/minivans, and run by your local transit agency instead of asshole bankers.
posted by aramaic at 2:08 PM on December 1, 2020
posted by aramaic at 2:08 PM on December 1, 2020
Tesla's FSD circuit board uses 2 independently powered chips which feed data to the part of the logic that makes the decision about what the car should do. If the chips don't agree on the content of a frame then that frame is thrown out.
Question. If the frame is the view of reality as determined by the sensors, and the processors are looking at the same frame, then if they don’t agree, isn’t the issue with the processors and not the sensors? Why is the frame discarded? How does this correct the issue? Does anything ask why there is an issue? #1 sees X and #2 sees Y when looking at the same picture then where lies the fault? Do we assume they both have the same exact frame to ponder? The more I think about the issues of automated driving the more my brain starts to hurt. Living in San Francisco, I see these prototype cars driving all over my neighborhood, I guess to learn something. I always check to see if the person in the drivers seat has their hands on the steering wheel. They always do. Thank you Jeebus.
posted by njohnson23 at 3:08 PM on December 1, 2020 [2 favorites]
Question. If the frame is the view of reality as determined by the sensors, and the processors are looking at the same frame, then if they don’t agree, isn’t the issue with the processors and not the sensors? Why is the frame discarded? How does this correct the issue? Does anything ask why there is an issue? #1 sees X and #2 sees Y when looking at the same picture then where lies the fault? Do we assume they both have the same exact frame to ponder? The more I think about the issues of automated driving the more my brain starts to hurt. Living in San Francisco, I see these prototype cars driving all over my neighborhood, I guess to learn something. I always check to see if the person in the drivers seat has their hands on the steering wheel. They always do. Thank you Jeebus.
posted by njohnson23 at 3:08 PM on December 1, 2020 [2 favorites]
aramaic: The issue I have with that is that it is heavily dependent on people having smart phones to summon the bus. This is yet another self-driving car issue: transportation inequity. Those who can afford it get to ride in their little self-driving car bubble, those who can't get unreliable mass transit.
posted by SansPoint at 3:10 PM on December 1, 2020 [3 favorites]
posted by SansPoint at 3:10 PM on December 1, 2020 [3 favorites]
...it is heavily dependent on people having smart phones...
Simple: give low-income people free smartphones with free data. California does. No reason other states/Feds couldn't do the same, which has the advantage of also giving them access to innumerable other digital services.
posted by aramaic at 3:57 PM on December 1, 2020 [1 favorite]
Simple: give low-income people free smartphones with free data. California does. No reason other states/Feds couldn't do the same, which has the advantage of also giving them access to innumerable other digital services.
posted by aramaic at 3:57 PM on December 1, 2020 [1 favorite]
That doesn't help people who don't know how to use them, or choose not to for whatever reason.
posted by SansPoint at 4:02 PM on December 1, 2020
posted by SansPoint at 4:02 PM on December 1, 2020
True, and it also doesn't help anyone living at the bottom of the ocean!
posted by aramaic at 4:04 PM on December 1, 2020 [8 favorites]
posted by aramaic at 4:04 PM on December 1, 2020 [8 favorites]
Fully automated driving seems like a foregone conclusion for freeways and especially freight. I don’t think it has a viable future inside cities, where transportation planners are biasing toward pedestrian, cycling, and transit uses that all make autonomy significantly harder by putting more obstacles and distractions in the road to get cars moving slower.
A parallel argument about a hot topic would have been flagged so fast. Consider the autonomy of those people biking, walking, etc. How is their personal autonomy limited by motor vehicles? Does the most dangerous user on the road really deserve the most autonomy? Why or why not?
posted by aniola at 4:13 PM on December 1, 2020 [4 favorites]
A parallel argument about a hot topic would have been flagged so fast. Consider the autonomy of those people biking, walking, etc. How is their personal autonomy limited by motor vehicles? Does the most dangerous user on the road really deserve the most autonomy? Why or why not?
posted by aniola at 4:13 PM on December 1, 2020 [4 favorites]
Foosnark. on making cars more efficient: Bikes are among the most (if not the most ) efficient vehicles. Using a ton of metal to move 1-2 humans around, not so much. The most efficient car is an electric bike. But the transportation infrastructure sure is designed for cars! You just described how efficiently cars are able to use the public space! It is time to optimize for the more vulnerable (and truly efficient) users of the public right of way.
posted by aniola at 4:13 PM on December 1, 2020 [4 favorites]
posted by aniola at 4:13 PM on December 1, 2020 [4 favorites]
aramaic: Nobody lives at the bottom of the ocean. Many people live without smart phones—by choice.
Transit is for everyone.
posted by SansPoint at 4:25 PM on December 1, 2020 [3 favorites]
Transit is for everyone.
posted by SansPoint at 4:25 PM on December 1, 2020 [3 favorites]
There are a lot of valid reasons one might not not use a smartphone. Valuing your own privacy. Respecting the privacy of others. Living within 100 miles of the US border. Money. Wrist Pain. Valuing your own time. Prioritizing creativity. These are some examples from my own personal list of reasons I don't use a smartphone.
I mean, I know I'm walking across the ocean right now with my fellow pirates (thanks for the setup) but I'd still hope transportation planners would care about me and the many other people without smartphones.
posted by aniola at 4:25 PM on December 1, 2020
I mean, I know I'm walking across the ocean right now with my fellow pirates (thanks for the setup) but I'd still hope transportation planners would care about me and the many other people without smartphones.
posted by aniola at 4:25 PM on December 1, 2020
On-demand transit is a thing in many places and so far the results have been pretty negative, as in way more expensive than just subsidizing a third party like taxis or Uber to provide the service. Fixed route has real advantages for people, even if they are mostly psychological.
posted by The_Vegetables at 4:56 PM on December 1, 2020 [2 favorites]
posted by The_Vegetables at 4:56 PM on December 1, 2020 [2 favorites]
A parallel argument about a hot topic would have been flagged so fast. Consider the autonomy of those people biking, walking, etc. How is their personal autonomy limited by motor vehicles? Does the most dangerous user on the road really deserve the most autonomy? Why or why not?Sorry, to clarify I was using the term “autonomy” in the sense of “autonomous driving” like the title of this post; unsure what the flag would be for. I don’t think the technology will be sufficiently good in our lifetimes to keep up with new street designs and more multi-modal users competing for road space. Smart transportation planners with adequate political air cover will prioritize single-occupant vehicles last.
posted by migurski at 5:42 PM on December 1, 2020
Fully autonomous cars have been 5 - 10 years away for at least 15 years now and I'd be shocked if they were actually working reliably by even 2030.
You know, we do have things like the Internet Archive, so you could link to examples. I'm not saying that it's impossible there were people in 2000 predicting fully autonomous cars by 2005 but I'd be interested to see who said that and if they also advised investing in Pets.com.
(Corporate motto: “Because Pets Can't Drive”, briefly remembered as one of the “greatest dotcom disasters”, opened November 1998 and according to Wikipedia probably raised ~$20 million, then spent $12 million on advertising alone in its first year, $11 million buying its biggest competitor, $1.2 million on just a Superb Owl ad in early 2000, and went bust by November of that year, but happily for Hasbro did not appear in the simultaneously-released Monopoly .com edition.)
scruss's William Huskisson/victim blaming bit reminds me of something I read years ago, which I'd swear was on MeFi but I can't find it, about early-twentieth-century P.R. campaigns by U.S. railroad companies to try to shape a public default assumption that if people were killed at road/railroad crossings it was the automobile driver's fault, featuring a black-and-white film of a car full of reckless teenagers who die horribly because of their terrible, terrible carelessness.
posted by XMLicious at 5:47 PM on December 1, 2020 [1 favorite]
We have had a solution in hand for 100 years (density! Trains! Busses!) but refuse to use it, basically because white people don’t want to share space with black people.
I don't disagree that cars are destroying cities, but one need look no farther than Miami to see that it isn't just racism that keeps transit from being successful in this country. Well, maybe it is still racism, I'm not in a good position to say, but I can conclude that it isn't just WASPs standing in the way of effective transit.
posted by wierdo at 6:34 PM on December 1, 2020
I don't disagree that cars are destroying cities, but one need look no farther than Miami to see that it isn't just racism that keeps transit from being successful in this country. Well, maybe it is still racism, I'm not in a good position to say, but I can conclude that it isn't just WASPs standing in the way of effective transit.
posted by wierdo at 6:34 PM on December 1, 2020
I am less of a self-driving car skeptic than most, but here's a transcript of a panel discussion at a major AI conference in 1984 with people worrying about AI hype getting out of control.
The very first thing they mention, way back then, is "suppose that five years from now the strategic computing initiative collapses miserably as autonomous vehicles fail to roll". So self-driving car hype has gone with AI hype for a long time, and even back then the timeline was 5 years.
posted by vogon_poet at 6:37 PM on December 1, 2020 [1 favorite]
The very first thing they mention, way back then, is "suppose that five years from now the strategic computing initiative collapses miserably as autonomous vehicles fail to roll". So self-driving car hype has gone with AI hype for a long time, and even back then the timeline was 5 years.
posted by vogon_poet at 6:37 PM on December 1, 2020 [1 favorite]
At the margin, these things may already be saving pedestrian, bicyclist, and passenger lives. At the margin, it's not a question of whether these machines drive better than the average driver, but whether these machines drive better than the average driver who buys this kind of $50-100K sports sedan. Watching the "enthusiastic reviews" video, with its beautiful documentation of about 100 driving errors in 20 minutes, I'd say that even with all that, the car is still doing a better, safer job than perhaps 90% of BMW drivers and 80% of Audi drivers I see on the road. It is often cautious and even hesitant when making turns, at least occasionally gives bicyclists and pedestrians space when passing them, and sometimes lets other cars go. None of those things would I say of the average BMW driver. If for the moment, at the margin, these cars are replacing only the assholes who drive expensive sports sedans, it may already be improving the world. Which of course says more about humans than the wonders of this technology.
posted by chortly at 6:40 PM on December 1, 2020 [3 favorites]
posted by chortly at 6:40 PM on December 1, 2020 [3 favorites]
The very first thing they mention, way back then, is "suppose that five years from now the strategic computing initiative collapses miserably as autonomous vehicles fail to roll".
In 1984, autonomous vehicles were focused more on military uses (think things like drones and gps guided missiles (as opposed to laser guided, wire guided, and dropped bombs) than self-driving cars for personal use. The accurate guided missiles, with video relay, featured in Bush #1's Gulf War were amazing uses of technology to the average person.
posted by The_Vegetables at 7:29 PM on December 1, 2020 [1 favorite]
In 1984, autonomous vehicles were focused more on military uses (think things like drones and gps guided missiles (as opposed to laser guided, wire guided, and dropped bombs) than self-driving cars for personal use. The accurate guided missiles, with video relay, featured in Bush #1's Gulf War were amazing uses of technology to the average person.
posted by The_Vegetables at 7:29 PM on December 1, 2020 [1 favorite]
Also, agreed on real public transit and high speed trains and denser cities,
Naturally the people in charge of doing that redevelopment will be the same idealistic people who put the freeway routes through our major cities. Undoubtedly the people in the communities that will be divided by new train tracks will be cheerfully enthusiastic about having their neighborhoods divvied up for the greater good.
From experience in various California towns, train routes are very effective methods of splitting off disagreeable neighborhoods. With attendant law enforcement opportunities
Team "fuck this noise" here. Give us buses, trains, electric bicycles and walkable, dense cities. Society has progressed past the need for cars.
Making cities dense and walkable is also a perfect excuse to gentrify neighborhoods. Take that unisightly neighborhood in say, Oakland or Atlanta, and turn it into upscale dense condos, along with a high-end chain stores and restaurants mix. Like Santana Row in San Jose, for example. All you need to do is get rid of the undesirable locals...
There have been a couple of interesting experiments on dynamically-dispatched buses, where the bus route depends on who's onboard, and the schedule depends on how many people are asking for buses.
That IS interesting. A natural development to that will be to dynamically dispatch buses based on income level, rider zip code, arrest and credit record, and so on. After all, the argument would go, wouldn't it be more important to devote resources to those who contribute the most in taxes?
posted by happyroach at 7:56 PM on December 1, 2020
Naturally the people in charge of doing that redevelopment will be the same idealistic people who put the freeway routes through our major cities. Undoubtedly the people in the communities that will be divided by new train tracks will be cheerfully enthusiastic about having their neighborhoods divvied up for the greater good.
From experience in various California towns, train routes are very effective methods of splitting off disagreeable neighborhoods. With attendant law enforcement opportunities
Team "fuck this noise" here. Give us buses, trains, electric bicycles and walkable, dense cities. Society has progressed past the need for cars.
Making cities dense and walkable is also a perfect excuse to gentrify neighborhoods. Take that unisightly neighborhood in say, Oakland or Atlanta, and turn it into upscale dense condos, along with a high-end chain stores and restaurants mix. Like Santana Row in San Jose, for example. All you need to do is get rid of the undesirable locals...
There have been a couple of interesting experiments on dynamically-dispatched buses, where the bus route depends on who's onboard, and the schedule depends on how many people are asking for buses.
That IS interesting. A natural development to that will be to dynamically dispatch buses based on income level, rider zip code, arrest and credit record, and so on. After all, the argument would go, wouldn't it be more important to devote resources to those who contribute the most in taxes?
posted by happyroach at 7:56 PM on December 1, 2020
Santana Row in San Jose was a run-down one-story outdoor shopping mall before; it’s not a case of gentrification. Same in Oakland for the most part where the absolute majority of dense residential developments are replacing parking lots or brownfields. This is especially true in the downtown area where Jerry Brown’s goal of 10,000 units was focused.
You’re right to be unhappy with gentrification, but dense housing and development are not the culprits you think they are!
posted by migurski at 8:08 PM on December 1, 2020 [7 favorites]
You’re right to be unhappy with gentrification, but dense housing and development are not the culprits you think they are!
posted by migurski at 8:08 PM on December 1, 2020 [7 favorites]
I don't really see autonomy reducing the number of cars on the road; rather I bet the number of trips/miles sky rockets as people no longer have to park or idle their cars. A lot of people who are incentivized to take transit because of the cost/availability of parking are going to be able to just put their cars in holding pattern mode. Rush hours could end up encompassing the majority of the day in ever more places.
I'd love to see autonomous cars rigidly adhere to speed limits though; even when the human is in charge. Neighbourhoods could arrange for really low limits (like 20km/h) for residential streets and actually have them be effective. As it stands now the playground zone in front of my house is just ignored by 50+% of drivers and it is infuriating.
It might greatly reduce the incentive to have a car capable of 160+km/h and the temptation to use that capability on public roads.
It would also be nice if these cars refused to park in bike lanes.
posted by Mitheral at 9:43 PM on December 1, 2020 [3 favorites]
I'd love to see autonomous cars rigidly adhere to speed limits though; even when the human is in charge. Neighbourhoods could arrange for really low limits (like 20km/h) for residential streets and actually have them be effective. As it stands now the playground zone in front of my house is just ignored by 50+% of drivers and it is infuriating.
It might greatly reduce the incentive to have a car capable of 160+km/h and the temptation to use that capability on public roads.
It would also be nice if these cars refused to park in bike lanes.
posted by Mitheral at 9:43 PM on December 1, 2020 [3 favorites]
As it stands now the playground zone in front of my house is just ignored by 50+% of drivers and it is infuriating.
Street design solves this problem quite effectively. The street I live on has parallel parking bays where there ought to be grass between the sidewalk and the street and few street trees, allowing for super long sight lines. Many people just fly down the street without a care in the world.
One street over, where there are a ton of trees making almost a tunnel and there are no parking bays so the parked cars make the lanes quite narrow (but still wide enough for cars to squeeze by each other) it is a completely different story. People crawl down that block, which is good since it has a pocket park.
posted by wierdo at 10:36 PM on December 1, 2020 [3 favorites]
Street design solves this problem quite effectively. The street I live on has parallel parking bays where there ought to be grass between the sidewalk and the street and few street trees, allowing for super long sight lines. Many people just fly down the street without a care in the world.
One street over, where there are a ton of trees making almost a tunnel and there are no parking bays so the parked cars make the lanes quite narrow (but still wide enough for cars to squeeze by each other) it is a completely different story. People crawl down that block, which is good since it has a pocket park.
posted by wierdo at 10:36 PM on December 1, 2020 [3 favorites]
If for the moment, at the margin, these cars are replacing only the assholes who drive expensive sports sedans, it may already be improving the world. Which of course says more about humans than the wonders of this technology.
If we look at companies like Waymo - they already have truly autonomous vehicles - just ones running on very limited routes. To evolve, Waymo just needs expand its areas covered - they should be able to use solid evidence of success in earlier trials to persuade regulators as they go - slow but simple. For Tesla's FSD, the way forward is trickier. At present their drivers are affluent technophiles: people happy to shell out thousands of dollars for beta software because its fun to play about with.
There is an implicit promise to these customers: if and when FSD reaches level 5 autonomy and is approved for use - you just sign up your car to work for Tesla's network and wave it goodbye in the evening as it goes off for a night of earning you commission by ferrying passengers about. The problem is that, to win over regulatory approval and general confidence of all concerned, there needs to be a tedious intermediate stage of "accompanied autonomy" : the car should have evolved to a point where it can be autonomous but we still need to have a driver at the steering wheel waiting to take over - even while we are starting to pick up passengers. I'm going to suggest that current Tesla owners are not going to volunteer to be that person - most especially if it is their car, their time and their liability if everything goes wrong.
Just a Waymo did, Tesla probably needs to employ a small army of such drivers and sit them in Tesla's own cars with Tesla taking responsibility should anything go wrong.
posted by rongorongo at 1:06 AM on December 2, 2020
If we look at companies like Waymo - they already have truly autonomous vehicles - just ones running on very limited routes. To evolve, Waymo just needs expand its areas covered - they should be able to use solid evidence of success in earlier trials to persuade regulators as they go - slow but simple. For Tesla's FSD, the way forward is trickier. At present their drivers are affluent technophiles: people happy to shell out thousands of dollars for beta software because its fun to play about with.
There is an implicit promise to these customers: if and when FSD reaches level 5 autonomy and is approved for use - you just sign up your car to work for Tesla's network and wave it goodbye in the evening as it goes off for a night of earning you commission by ferrying passengers about. The problem is that, to win over regulatory approval and general confidence of all concerned, there needs to be a tedious intermediate stage of "accompanied autonomy" : the car should have evolved to a point where it can be autonomous but we still need to have a driver at the steering wheel waiting to take over - even while we are starting to pick up passengers. I'm going to suggest that current Tesla owners are not going to volunteer to be that person - most especially if it is their car, their time and their liability if everything goes wrong.
Just a Waymo did, Tesla probably needs to employ a small army of such drivers and sit them in Tesla's own cars with Tesla taking responsibility should anything go wrong.
posted by rongorongo at 1:06 AM on December 2, 2020
As someone who's been in software for a long time, all I can say is: I'm never getting into one of those things. No way, no how.
As co-founder of a company of ethical hackers (and having been in software a loooong time too), let me tell you you're not looking broadly enough. I refuse to get into any car that is connected to the internet, never mind any driving assistance software on top of that, autonomous or otherwise.
posted by DreamerFi at 5:52 AM on December 2, 2020 [3 favorites]
As co-founder of a company of ethical hackers (and having been in software a loooong time too), let me tell you you're not looking broadly enough. I refuse to get into any car that is connected to the internet, never mind any driving assistance software on top of that, autonomous or otherwise.
posted by DreamerFi at 5:52 AM on December 2, 2020 [3 favorites]
> As co-founder of a company of ethical hackers (and having been in software a loooong time too), let me tell you you're not looking broadly enough. I refuse to get into any car that is connected to the internet, never mind any driving assistance software on top of that, autonomous or otherwise.
Indeed, it seems like software people are if anything overrepresented in the autonomous car skeptic community in the same way that former line cooks are careful about where they dine out and what they eat. It's a lot more comforting to assume that the nerds have everything figured out, but it's hard to do that when you're one of the nerds who, if you had everything figured out, would be living in a beach house on an island somewhere with your own human driver to take you wherever you need to go.
posted by tonycpsu at 6:13 AM on December 2, 2020 [2 favorites]
Indeed, it seems like software people are if anything overrepresented in the autonomous car skeptic community in the same way that former line cooks are careful about where they dine out and what they eat. It's a lot more comforting to assume that the nerds have everything figured out, but it's hard to do that when you're one of the nerds who, if you had everything figured out, would be living in a beach house on an island somewhere with your own human driver to take you wherever you need to go.
posted by tonycpsu at 6:13 AM on December 2, 2020 [2 favorites]
Yeah, I'm a software test engineer who isn't happy about the idea of trusting my life to software developers every time I cross the street.
posted by octothorpe at 7:32 AM on December 2, 2020 [1 favorite]
posted by octothorpe at 7:32 AM on December 2, 2020 [1 favorite]
I'm reminded of the 737 Max problem too. A software system that was designed to help pilots was buggy, undocumented, and managed to kill hundreds of people and endanger who knows how many others, yet it was approved by the FAA because Boeing's people said it was safe—not because the FAA audited it to confirm it was safe.
So, to sum up, self-driving cars put too much trust in buggy software that is not being audited—just tested on streets. Testing that, I should add, already has a body count (thanks to Uber, of course). It still has nearly all the problems of cars: poor capacity, inefficient utilization of road infrastructure, and prioritizes cars over other modes of transportation.
We're solving the wrong problem, at the wrong level, in the wrong way, in a way that puts too many people at risk. Humans are shitty drivers. Fine. But that is only part of a gigantic set of problems around driving and car culture, and self-driving cars only fix one of those, it doesn't even fix it well, and it adds a whole new set of problems on top of it that are getting short-shrift in favor of nonsense about the Trolley Problem.
posted by SansPoint at 7:39 AM on December 2, 2020 [6 favorites]
So, to sum up, self-driving cars put too much trust in buggy software that is not being audited—just tested on streets. Testing that, I should add, already has a body count (thanks to Uber, of course). It still has nearly all the problems of cars: poor capacity, inefficient utilization of road infrastructure, and prioritizes cars over other modes of transportation.
We're solving the wrong problem, at the wrong level, in the wrong way, in a way that puts too many people at risk. Humans are shitty drivers. Fine. But that is only part of a gigantic set of problems around driving and car culture, and self-driving cars only fix one of those, it doesn't even fix it well, and it adds a whole new set of problems on top of it that are getting short-shrift in favor of nonsense about the Trolley Problem.
posted by SansPoint at 7:39 AM on December 2, 2020 [6 favorites]
I refuse to get into any car that is connected to the internet, never mind any driving assistance software on top of that, autonomous or otherwise.
Every new car sold in Europe since about 2018 is connected to the internet to support their 911 crash laws. Since the same cars are generally sold in the US, I wouldn't be surprised if they leave the tech in the cars when they are sold in the US, and trickle down says the same tech will be added in the US soon. When we were working with GM a few years ago, they had like 15 cellular antennas in their nicer cars.
posted by The_Vegetables at 7:57 AM on December 2, 2020
Every new car sold in Europe since about 2018 is connected to the internet to support their 911 crash laws. Since the same cars are generally sold in the US, I wouldn't be surprised if they leave the tech in the cars when they are sold in the US, and trickle down says the same tech will be added in the US soon. When we were working with GM a few years ago, they had like 15 cellular antennas in their nicer cars.
posted by The_Vegetables at 7:57 AM on December 2, 2020
There is an implicit promise to these customers: if and when FSD reaches level 5 autonomy and is approved for use - you just sign up your car to work for Tesla's network and wave it goodbye in the evening as it goes off for a night of earning you commission by ferrying passengers about.
Musk has been promising this for years and sucking in the money of gullible customers for FSD by the billions. He said that there would be a million robotaxis on the road by the end of 2020. He always says you better buy the full safe driving option now. And people believe him and hand over their money. He's a scam artist.
posted by JackFlash at 8:42 AM on December 2, 2020 [4 favorites]
Musk has been promising this for years and sucking in the money of gullible customers for FSD by the billions. He said that there would be a million robotaxis on the road by the end of 2020. He always says you better buy the full safe driving option now. And people believe him and hand over their money. He's a scam artist.
posted by JackFlash at 8:42 AM on December 2, 2020 [4 favorites]
Software failures are funny when the computer is just controlling a camera. Not so much when it's controlling a 2,000-4,000lb vehicle going fast enough to have a 50% chance of killing anyone it hits.
Seems like we're pretty much agreed that humans aren't very good drivers. I don't see much reason to expect human-designed technologies to be significantly safer - certainly not in a market-driven economy, and especially not in one where wealthy corporations have disproportionate influence over legislation.
Can we put this whole autonomous vehicles thing on the backburner and focus more energy on better driver education or infrastructure design or incentivising existing less-deadly modes of transportation and de-incentivising driving wherever it's feasible? Please?
posted by sibilatorix at 2:19 PM on December 2, 2020 [2 favorites]
Seems like we're pretty much agreed that humans aren't very good drivers. I don't see much reason to expect human-designed technologies to be significantly safer - certainly not in a market-driven economy, and especially not in one where wealthy corporations have disproportionate influence over legislation.
Can we put this whole autonomous vehicles thing on the backburner and focus more energy on better driver education or infrastructure design or incentivising existing less-deadly modes of transportation and de-incentivising driving wherever it's feasible? Please?
posted by sibilatorix at 2:19 PM on December 2, 2020 [2 favorites]
sibilatorix: Can we put this whole autonomous vehicles thing on the backburner and focus more energy on better driver education or infrastructure design or incentivising existing less-deadly modes of transportation and de-incentivising driving wherever it's feasible? Please?
Of course not. There's no money in that. /s
posted by SansPoint at 2:25 PM on December 2, 2020 [2 favorites]
Of course not. There's no money in that. /s
posted by SansPoint at 2:25 PM on December 2, 2020 [2 favorites]
autonomous vehicles -- the ultimate killer app
posted by philip-random at 2:44 PM on December 2, 2020
posted by philip-random at 2:44 PM on December 2, 2020
This discussion will be eventually be decided for us by the insurance industry. Autonomous cars are not perfect drivers yet, but they're already better than some imperfect human drivers. When they get better than most humans (and if they can crest the costs of some class action lawsuits, as opposed to the larger societal cost of ongoing and shrug-accepted drunk driver carnage) then the dynamic will flip to another chapter.
(I heart Europa-style public transit & intercity rail as much as anyone, but it's a pretty tough sell in the USA because: it's Socialism! Not visible on the radar yet, but if I was optimistic I'd mention some other progressive ideas that are gradually filtering into American culture, like gender rights and downgrading the war on drugs. More Socialism in USA was once unthinkable, now it's merely controversial, maybe in the future it will be possible.)
posted by ovvl at 3:58 PM on December 2, 2020 [3 favorites]
(I heart Europa-style public transit & intercity rail as much as anyone, but it's a pretty tough sell in the USA because: it's Socialism! Not visible on the radar yet, but if I was optimistic I'd mention some other progressive ideas that are gradually filtering into American culture, like gender rights and downgrading the war on drugs. More Socialism in USA was once unthinkable, now it's merely controversial, maybe in the future it will be possible.)
posted by ovvl at 3:58 PM on December 2, 2020 [3 favorites]
Elon Musk: "Mass transit is a pain in the ass. And there’s like a bunch of random strangers, one of who might be a serial killer. Individualized mass transit is the future."
Individualized mass transit, whatever the hell that is. And this is the guy people proclaim as the visionary for the future. He's a sociopath.
posted by JackFlash at 5:50 PM on December 2, 2020 [4 favorites]
Individualized mass transit, whatever the hell that is. And this is the guy people proclaim as the visionary for the future. He's a sociopath.
posted by JackFlash at 5:50 PM on December 2, 2020 [4 favorites]
fwiw...
on tech solutionism vs. political dysfunction/progress: "Politics and technology should not have to be substitutes. They should work together to solve problems. When you don't have one, you have to lean harder on the other, but that's far from an ideal situation."
posted by kliuless at 9:54 PM on December 2, 2020
on tech solutionism vs. political dysfunction/progress: "Politics and technology should not have to be substitutes. They should work together to solve problems. When you don't have one, you have to lean harder on the other, but that's far from an ideal situation."
posted by kliuless at 9:54 PM on December 2, 2020
The Anti-Software Crowd is, to me, mind boggling, especially when it comes from fellow software engineers and designers.
Many arguments read like NIMBYisms, or have a strong hint of "I've got mine" attitude.
For every "not everyone has a smartphone" there's a blind person, or a disabled person who can use a smartphone, but can't drive. Or lives in a village in a region with declining population, no local supermarket, without a personal driver, and no chance of decent public transportation, ever. But hey grandpa, move to London! Or Hong Kong, the choice is yours.
If you're categorically against software controlling moving hardware: have you really put yourself in someone else's shoes?
Then there's the "my software is a dirty kitchen, so everyone else's is too" thinking which I won't unpack here.
posted by romanb at 2:49 AM on December 3, 2020 [4 favorites]
Many arguments read like NIMBYisms, or have a strong hint of "I've got mine" attitude.
For every "not everyone has a smartphone" there's a blind person, or a disabled person who can use a smartphone, but can't drive. Or lives in a village in a region with declining population, no local supermarket, without a personal driver, and no chance of decent public transportation, ever. But hey grandpa, move to London! Or Hong Kong, the choice is yours.
If you're categorically against software controlling moving hardware: have you really put yourself in someone else's shoes?
Then there's the "my software is a dirty kitchen, so everyone else's is too" thinking which I won't unpack here.
posted by romanb at 2:49 AM on December 3, 2020 [4 favorites]
SansPoint: I find this interesting because every visualization of the self-driving car future treats pedestrians as an afterthought if it even thinks about them at all
A special shout Carol, potentially foolhardy wife of FSD beta tester AIDRIVR, who has been dressing in black on unlit roads and stepping into the path of her husbands car to see how it would react. The performance of the car in these kinds of events - whether staged or naturally occurring, get flagged as edge cases to Tesla. At present there is apparently a small army of people who are manually flagging and labeling these cases to try to improve the software. From next year Tesla plans to introduce a computer system called Dojo to do this more automatically and quickly.
At present the system is still failing some of these staged tests - and it is quitting FSD and handing over control to the driver in others: also a fail. However - with sensory capabilities , powers of attention, reaction times and a knowledgeable of experience that all exceed human drivers: I'd put my money on the performance exceeding that that of people within a few months. There are other technical problems with autonomous system may struggle with - but avoiding and respecting pedestrians should not be one of them, IMHO.
(When it comes to avoiding hitting people then autonomous systems will sometimes face nightmarish dilemmas: do they swerve to avoid a pedestrian if that will take the car straight into the path of an oncoming vehicle, for example? We expect human drivers to make the same choices but understand the pressure they are under. Autonomous systems will retain the logic by which they made a choice).
posted by rongorongo at 4:53 AM on December 3, 2020
A special shout Carol, potentially foolhardy wife of FSD beta tester AIDRIVR, who has been dressing in black on unlit roads and stepping into the path of her husbands car to see how it would react. The performance of the car in these kinds of events - whether staged or naturally occurring, get flagged as edge cases to Tesla. At present there is apparently a small army of people who are manually flagging and labeling these cases to try to improve the software. From next year Tesla plans to introduce a computer system called Dojo to do this more automatically and quickly.
At present the system is still failing some of these staged tests - and it is quitting FSD and handing over control to the driver in others: also a fail. However - with sensory capabilities , powers of attention, reaction times and a knowledgeable of experience that all exceed human drivers: I'd put my money on the performance exceeding that that of people within a few months. There are other technical problems with autonomous system may struggle with - but avoiding and respecting pedestrians should not be one of them, IMHO.
(When it comes to avoiding hitting people then autonomous systems will sometimes face nightmarish dilemmas: do they swerve to avoid a pedestrian if that will take the car straight into the path of an oncoming vehicle, for example? We expect human drivers to make the same choices but understand the pressure they are under. Autonomous systems will retain the logic by which they made a choice).
posted by rongorongo at 4:53 AM on December 3, 2020
I can't help but think that if Autonomous-car Software A can be trained to cut off in traffic / never yield to Autonomous-car Software B, that that is a thing that is going to happen. Because people are assholes.
Anyone think Musk wouldn't try to get away with something like that? Because if he can advertise "No autonomous Tesla will ever let an Other Car Company car get in front of it", that would lead to lots of sales and I don't think he would give a fuck about the backlash. Not with his ego. Especially if Other Car Company then advertises "We will take on Teslas and not yield to them". Then what? Is anything like this anticipated by the regulators?
And I think there are just too many edge-cases. Like they may train these things to deal with bicycles ok, but what if a bicyclist swerves around a puddle? And wasn't one of the deaths from an autonomous system of a person who was walking a bicycle across a crosswalk, a situation not anticipated / programmed for by the software designers?
What about unicycles? Those exist. And the super-tall extended unicycles that one might find a juggler on top of? What about tricycles? (yes, they make full-sized ones as well as child-sized ones). What about people towing trailers with a bike, carrying children or other things? What about tandem bikes? Tandem bikes towing trailers? What about those old-timey-style bikes with the huge front wheel? What about drunk bicyclists or pedestrians? These are just off the top of my head, there are certainly many, many more.
And the question of "who is liable?", has that been answered?
I think when the software is so complicated that it becomes a black box, the makers can go "Well we didn't *explicitly* program it to kill those people, so we are not liable", or "Well it was impossible to anticipate that the Swerve Handling Sub-Module would interact with the Abrupt Stopping Sub-Module in that way in that edge-case situation, because we had never encountered it in testing, and we shouldn't be expected to handle all possible cases, that would be ridiculous, so we are not liable". And the manufacturers will say the same thing. And the car insurer will say the same thing. Who, then, is liable? And whoever it is, they better have deep pockets, because once autonomous vehicles are widespread, this shit will be happening all the time.
posted by cats are weird at 5:50 AM on December 3, 2020 [2 favorites]
Anyone think Musk wouldn't try to get away with something like that? Because if he can advertise "No autonomous Tesla will ever let an Other Car Company car get in front of it", that would lead to lots of sales and I don't think he would give a fuck about the backlash. Not with his ego. Especially if Other Car Company then advertises "We will take on Teslas and not yield to them". Then what? Is anything like this anticipated by the regulators?
And I think there are just too many edge-cases. Like they may train these things to deal with bicycles ok, but what if a bicyclist swerves around a puddle? And wasn't one of the deaths from an autonomous system of a person who was walking a bicycle across a crosswalk, a situation not anticipated / programmed for by the software designers?
What about unicycles? Those exist. And the super-tall extended unicycles that one might find a juggler on top of? What about tricycles? (yes, they make full-sized ones as well as child-sized ones). What about people towing trailers with a bike, carrying children or other things? What about tandem bikes? Tandem bikes towing trailers? What about those old-timey-style bikes with the huge front wheel? What about drunk bicyclists or pedestrians? These are just off the top of my head, there are certainly many, many more.
And the question of "who is liable?", has that been answered?
I think when the software is so complicated that it becomes a black box, the makers can go "Well we didn't *explicitly* program it to kill those people, so we are not liable", or "Well it was impossible to anticipate that the Swerve Handling Sub-Module would interact with the Abrupt Stopping Sub-Module in that way in that edge-case situation, because we had never encountered it in testing, and we shouldn't be expected to handle all possible cases, that would be ridiculous, so we are not liable". And the manufacturers will say the same thing. And the car insurer will say the same thing. Who, then, is liable? And whoever it is, they better have deep pockets, because once autonomous vehicles are widespread, this shit will be happening all the time.
posted by cats are weird at 5:50 AM on December 3, 2020 [2 favorites]
What about unicycles?
There are many white papers on this topic, and a number of solutions. If you flip the logic around, your problem becomes simpler, at least in theory. You don't need to classify each and every possible potentially mobile object ("a bird riding a horse with two heads", "a pair of donkeys pulling a sled full of cacti"). Software can treat any object as something that can unexpectedly move in front of a vehicle. Say, a tree that is falling onto the road because it was struck by lightning. It does not matter that it is a tree; it does matter that an object is on the road in front of the car. The software does not need to "know" what a tandem with a trailer is.
That's not to say that there are no edge cases.
posted by romanb at 7:14 AM on December 3, 2020 [1 favorite]
There are many white papers on this topic, and a number of solutions. If you flip the logic around, your problem becomes simpler, at least in theory. You don't need to classify each and every possible potentially mobile object ("a bird riding a horse with two heads", "a pair of donkeys pulling a sled full of cacti"). Software can treat any object as something that can unexpectedly move in front of a vehicle. Say, a tree that is falling onto the road because it was struck by lightning. It does not matter that it is a tree; it does matter that an object is on the road in front of the car. The software does not need to "know" what a tandem with a trailer is.
That's not to say that there are no edge cases.
posted by romanb at 7:14 AM on December 3, 2020 [1 favorite]
And I think there are just too many edge-cases.
Those edge cases currently exist, and tend to result in death, usually with the victim being blamed.
...but also, what happens if both vehicle video systems fail simultaneously, while the would-be driver is asleep, and also that driver is deaf so they can't hear the warning siren, and it's also a hailstorm, and then two babies bicycle across the street at that moment while wearing white, does the car drive over the babies, or steer into the onrushing tornado, knowing if it does that the car will be whisked into the air and potentially dropped on to a bus full of kindergartners? What then? What then?
posted by aramaic at 7:26 AM on December 3, 2020 [7 favorites]
Those edge cases currently exist, and tend to result in death, usually with the victim being blamed.
...but also, what happens if both vehicle video systems fail simultaneously, while the would-be driver is asleep, and also that driver is deaf so they can't hear the warning siren, and it's also a hailstorm, and then two babies bicycle across the street at that moment while wearing white, does the car drive over the babies, or steer into the onrushing tornado, knowing if it does that the car will be whisked into the air and potentially dropped on to a bus full of kindergartners? What then? What then?
posted by aramaic at 7:26 AM on December 3, 2020 [7 favorites]
Serious question: what if i wear clothing designed to protect me from appearing as a human to surveillance cameras? I foresee purposefully obfuscating designs as becoming more popular.
posted by tofu_crouton at 8:34 AM on December 3, 2020 [1 favorite]
posted by tofu_crouton at 8:34 AM on December 3, 2020 [1 favorite]
What about unicycles?
Short piece of footage taken today of FSD detecting and avoiding a deer in the same manner as it would a pedestrian. It may be that the system does not distinguish deer, pedestrian or uni-cyclist in the road - and treats them all the same "Gaussian potential field" in the word of romanb's paper. They are all really just obstacles that need to be noticed and avoided.
posted by rongorongo at 11:26 AM on December 3, 2020
Short piece of footage taken today of FSD detecting and avoiding a deer in the same manner as it would a pedestrian. It may be that the system does not distinguish deer, pedestrian or uni-cyclist in the road - and treats them all the same "Gaussian potential field" in the word of romanb's paper. They are all really just obstacles that need to be noticed and avoided.
posted by rongorongo at 11:26 AM on December 3, 2020
What about unicycles? II
For anybody interested in the technicalities how Tesla's FSD is learning edge cases - I'd recommend this talk between Dave Lee and deep learning researcher James Douma.
They talk about a sequence which might be slightly counter-intuitive and which goes like this:
1. Some chairs fall off the back of a truck onto the road. A Tesla, which knows nothing about this sort of event, is following the truck. Bad things happen - a crash or near crash - but the camera footage survives.
2. Tesla engineers choose to teach FSD about this edge case. But to do that they need lots of footage of chairs - or other stuff, falling off trucks.
3. Happily they can reach out to the whole fleet of cars to request that footage. They look at the pattern of information derived from the falling chair video footage and they ask all cars in the fleet to look out for anything similar and upload any close matches.
4. After a while they get the set of footage they need to label the footage train the neural network about how to best avoid incidents of this kind - they then push that knowledge out to the network as part of an update.
- So that it how the system would deal with unicyclists, if they provide to have a meaningfully different behaviour profile from other cyclists.
posted by rongorongo at 4:43 AM on December 5, 2020 [2 favorites]
For anybody interested in the technicalities how Tesla's FSD is learning edge cases - I'd recommend this talk between Dave Lee and deep learning researcher James Douma.
They talk about a sequence which might be slightly counter-intuitive and which goes like this:
1. Some chairs fall off the back of a truck onto the road. A Tesla, which knows nothing about this sort of event, is following the truck. Bad things happen - a crash or near crash - but the camera footage survives.
2. Tesla engineers choose to teach FSD about this edge case. But to do that they need lots of footage of chairs - or other stuff, falling off trucks.
3. Happily they can reach out to the whole fleet of cars to request that footage. They look at the pattern of information derived from the falling chair video footage and they ask all cars in the fleet to look out for anything similar and upload any close matches.
4. After a while they get the set of footage they need to label the footage train the neural network about how to best avoid incidents of this kind - they then push that knowledge out to the network as part of an update.
- So that it how the system would deal with unicyclists, if they provide to have a meaningfully different behaviour profile from other cyclists.
posted by rongorongo at 4:43 AM on December 5, 2020 [2 favorites]
More information is trickling out about Waymo, making for a brief and reasonably interesting article at Ars Technica, in the course of which they mention a YouTube channel with a bunch of ride videos.
posted by aramaic at 8:39 AM on December 7, 2020 [1 favorite]
posted by aramaic at 8:39 AM on December 7, 2020 [1 favorite]
Uber’s fraught and deadly pursuit of self-driving cars is over The ride-hailing company will sell its Advanced Technologies Group to Aurora Innovation
posted by octothorpe at 1:19 PM on December 7, 2020 [1 favorite]
posted by octothorpe at 1:19 PM on December 7, 2020 [1 favorite]
NIkki from Transport Evolved provides a nice summary of where we are with autonomous vehicles at the end of 2020, why things are taking longer than promised and about the uncanny valley to be crossed:
Elon Musk, Others, Promised Us Autonomous Cars By Now. Where Are They?
posted by rongorongo at 4:09 AM on December 16, 2020 [2 favorites]
Elon Musk, Others, Promised Us Autonomous Cars By Now. Where Are They?
posted by rongorongo at 4:09 AM on December 16, 2020 [2 favorites]
« Older A US immigrant on racism and shame | Nut Cracker Suite Newer »
This thread has been archived and is closed to new comments
For those interested in Tesla's AI approach: previously their FSD module polled the vehicle's 8 cameras, radar and sonar sensors about 30 times a second to produce frames which were then analysed. This worked after a fashion but was slow and error prone. The re-written software integrates all the camera feeds into one - and then analyses them as continuous video rather than discrete images. It can perform "trajectory projection" - working out whether an object which it saw but which has momentarily disappeared from view - poses a threat, for example. The system is also capable of improving and learning in various ways.
The is my second attempt at this FPP - my first was rather wide-eyed "look at this technology" post. But it is difficult to see the promises of the technology in isolation from its dangers - the mods deleted it and asked me to try for something more balanced.
posted by rongorongo at 7:09 AM on December 1, 2020 [13 favorites]