Cory Doctorow: 'Self-Driving Cars are Bullshit' (pluralistic.net) 347
"Self-driving cars are bullshit," writes Cory Doctorow:
I'm a science fiction writer, so I quite enjoy thinking about self-driving cars. They make for really interesting analogies about data, liability, self-determination, information security and openness... But I'm a science fiction writer and that means I can tell the difference between "thought experiments" and "real things." Alas, the same cannot be said of corporate America.
For example, according to its own IPO filings, Uber can only be profitable if it invents fully autonomous vehicles and replaces every public transit ride in the world with them. Elon Musk — a man whose "green electric car company" is only profitable thanks to the carbon credits it sells to manufacturers of the dirtiest SUVs in America, without which those planet-killing SUVs would not exist — makes the same mistake. Musk wants to abolish public transit and replace it with EVs (he says that public transit makes you sit next to strangers who might be serial killers, which tells you a lot about his view of humanity).
Now, both Uber and Musk are both wrong as a matter of simple geometry. Multiply the space occupied by all those AVs by the journeys people in cities need to make by the additional distances of those journeys if we need road for all those cars, and you run out of space... these fairy tales require so much credulity to be taken seriously that they strain even the car-addled imaginations of American automotive culture, and also rely on the irrational exuberance inspired by imaginary self-driving cars to propagate and persist.
But that exuberance is sorely misplaced. Machine learning systems have brittle and unpredictable failure modes that can be triggered by accident or deliberately. The unconstrained problem of navigating busy cities with unquantifiable human activities is insoluble with ML.
Or, at least, it's insoluble if you care about whether cars kill even more people in even less predictable ways than they do now...
Doctorow adds that a key plot point in the third book in his "Little Brother" series (coming out in October) is "subverted, lethal autonomous vehicles." But Doctorow also shares a link to his short story "Car Wars," commissioned by Deakin University to explore the sociotechnological issues around autonomous vehicle.
It begins with a high school warning parents about students performing "dangerous modifications" to their car in violation of new federal laws -- three student vehicles were already confiscated for "operating with unlicensed firmware." (And "one of those cases has been referred to the police as the student involved was a repeat offender.")
But as the school launches its random vehicle firmware audits, a developer sends a desperate message to his followers on Twitter about something even more disturbing that's happening in real-time...
For example, according to its own IPO filings, Uber can only be profitable if it invents fully autonomous vehicles and replaces every public transit ride in the world with them. Elon Musk — a man whose "green electric car company" is only profitable thanks to the carbon credits it sells to manufacturers of the dirtiest SUVs in America, without which those planet-killing SUVs would not exist — makes the same mistake. Musk wants to abolish public transit and replace it with EVs (he says that public transit makes you sit next to strangers who might be serial killers, which tells you a lot about his view of humanity).
Now, both Uber and Musk are both wrong as a matter of simple geometry. Multiply the space occupied by all those AVs by the journeys people in cities need to make by the additional distances of those journeys if we need road for all those cars, and you run out of space... these fairy tales require so much credulity to be taken seriously that they strain even the car-addled imaginations of American automotive culture, and also rely on the irrational exuberance inspired by imaginary self-driving cars to propagate and persist.
But that exuberance is sorely misplaced. Machine learning systems have brittle and unpredictable failure modes that can be triggered by accident or deliberately. The unconstrained problem of navigating busy cities with unquantifiable human activities is insoluble with ML.
Or, at least, it's insoluble if you care about whether cars kill even more people in even less predictable ways than they do now...
Doctorow adds that a key plot point in the third book in his "Little Brother" series (coming out in October) is "subverted, lethal autonomous vehicles." But Doctorow also shares a link to his short story "Car Wars," commissioned by Deakin University to explore the sociotechnological issues around autonomous vehicle.
It begins with a high school warning parents about students performing "dangerous modifications" to their car in violation of new federal laws -- three student vehicles were already confiscated for "operating with unlicensed firmware." (And "one of those cases has been referred to the police as the student involved was a repeat offender.")
But as the school launches its random vehicle firmware audits, a developer sends a desperate message to his followers on Twitter about something even more disturbing that's happening in real-time...
Not only are they bullshit, but they're junk (Score:2)
Half baked contraptions, like computers themselves.
That is not to say that they won't be far superior to humans. If you don't believe me, look at the safety records of airlines. Unless it's done intentionally, they are not flying into mountains like they regularly used to.
Re: (Score:2)
http://avherald.com/h?article=4db42eb9
Re:Not only are they bullshit, but they're junk (Score:4, Insightful)
If I read that correctly, it was a CFIT accident involving a fairly old, small, low-tech aircraft owned by a company that was shut down for having too many crashes, leased to somebody else, and flown using GPS that isn't certified for aircraft use, which crashed into a mountain because the altitude of the mountain in the GPS system was wrong. It's amazing how many corners were cut there....
Re:Not only are they bullshit, but they're junk (Score:4, Insightful)
No, thanks.
Re:Not only are they bullshit, but they're junk (Score:4, Insightful)
Yeah, actually it was the corrupt human element that certified an unairworthy aircraft to save a penny. The cause is obvious, human greed. Properly designed and maintained machines don't have that problem.
And note those were the only fatal accidents caused by deficient design in years. The rest were pilot error (intentional and otherwise) and shoot downs.
Airlines are many times safer than just 15 years ago. Humans are the biggest danger by far. And if the machine ever recognize that, you know what will happen!
Re: (Score:2)
And look at how many air accidents of the ones mentioned were literally cases of CFIT - computerized flight into terrain, caused by pilot over-reliance of faulty computer sensors over common sense
Re: Not only are they bullshit, but they're junk (Score:3)
With 2 million AV capable vehicles in the road driving 30 miles a day that's 1.8 Billion miles of validation per month.
If engineers want to confirm safety all they have to do is run it in user supervised mode for 6 months and I can't imagine a governing agency in the world declining technology with 10+ Billion miles of verification.
That's completely different from the 737 Max in which each failure kills hundreds and the sample size for testing is relatively small.
Re: (Score:3)
Re: (Score:2)
Yeah, you couldn't get 1,000 people into three 737 MAX aircraft, or even close (690 maximum, even with the biggest model, in its densest configuration, which the two planes that crashed were not).
Re: Not only are they bullshit, but they're junk (Score:2)
Makes no difference. There were defects at every level in the design and implementation.
The numbers of deaths don't count, because that's random. What matters is the avoidable violation of methodology count and the avoidable defect count that results.
Re:Not only are they bullshit, but they're junk (Score:5, Insightful)
Very well put. To many people, "self-driving car" and "AI" are one and the same, and since they believe we can never have "AI", it must therefore be true that we can never have "self-driving cars". But that ignores the entire history of AI research.
A problem is called "AI-complete" when it seems like we'd need (strong) AI to solve it. However, there's a decades-old joke in the field that (strong) AI is the set of AI-complete problems that haven't been solved yet. Speech recognition and computer vision (specifically picking simple objects out from an image) are two solved problems that we long considered AI-complete. Note that "solved" here doesn't mean "perfect"! Humans also make mistakes in these things. It means: works well enough to make products. As soon as each such problem is solved, the goalposts are moved, they're no longer an important part of (strong) AI, and (strong) AI is still impossible, along with all AI-complete problems.
Clearly self-driving cars fit this pattern. As soon as we have acceptably-working products, all the AI purists will flip from "we can't have it because AI is impossible" to "it never had anything to do with AI, what nonsense, its solution means nothing" and we'll get on with life.
Self-driving cars are a hard problem. Humanity solves hard problems all the time. The timeframe may seem long by internet/software standards, but they've been coming along remarkably quickly by the standards of innovation in the automobile industry, with its 5-year release cycles. Likely we're just a couple of such cycles away now. 10 years may be "never" in internet years, but it's right around the corner by industry standards.
Re: (Score:2)
And LIDAR, please! Rube Goldberg would be mighty proud of that one. And that's just for starters...
Remember when DARPA wanted to crowdsource a working product out of the LIDAR patents they bought, so they made that robot obstacle course, and they all failed miserably, and everyone was like "we'll never have robots, look at how they fail! We are so far from being able to invent good robots as a species!"?
Of course, the whole time, it was obvious that the shitty LIDAR itself that was the problem. Hilarious.
Re: (Score:3)
On in the case of machine learning: "When we figure out how it works, it stops being AI."
Re: (Score:3)
Re: (Score:3)
It's important to start small, then expand. (Score:5, Insightful)
First, there are places where autonomous vehicles would rock, such as airport shuttles running between gates and transportation facilities (lots, rentals, bus, trains, etc.). They would use existing road infrastructure and avoid the cost of light rail until it can be funded and built.
Also, the idea of "fully autonomous" is a bit of a misnomer: All such vehicles rely on their environment containing "clues". Rather than have all sensors be on the car and all clues be passive, some external clues can and should be active, perhaps embedded in the road or built into signs, so that the car vision/radar/sonar/lidar and sensing systems would have some help and redundancy.
There is a chunk of I-15 near me that has magnets embedded down the middle of several lanes. The magnet spacing is used to convey multiple types of information, such as speed, position in the lane, lane number, and are extremely reliable to read with a simple sensor under the vehicle. The magnets also identify when a lane is being added or removed, and if there's an HOV lane merge approaching.
All this from drilling small holes and dropping strong magnets into them. It's actually cheaper than mounting road reflectors and Bott's Dots (which California has recently stopped doing).
Let's use autonomous vehicles in closed areas (like airports). Let's make dumb roads just a little smarter.
Re: (Score:2)
Airports are the last place AVs will be useful. Most people are going in the same direction, the whole thing is laid out to make access to mass transit easy.
Re: (Score:2)
Let's use autonomous vehicles in closed areas (like airports). Let's make dumb roads just a little smarter.
Exactly. There plenty of places with routes that vehicles travel regularly, like airport shuttles. They follow one route back and forth or round and round all day long. That could easily be programmed in since the route is reliable and predictable.
Am I missing something here? (Score:5, Interesting)
His point seems to be more centered around objecting to the idea that public transit could somehow be replaced by fleets of self driving vehicles that are only (apparently) low-occupancy.
When in fact, public transit seems to me to offer the virtually perfect opportunity *FOR* autonomous vehicles, as not needing to pay salaries to drivers would enable public transit to efficiently service larger areas and with greater frequency, factoring in two very popular reasons why some are opposed to using public transit themselves (there are other reasons that are often cited, but these two still do come up a lot).
Re: (Score:2)
Re:Am I missing something here? (Score:5, Insightful)
He's also full of ignorant nonsense.
1) Tesla is far more *anti-incentivized* in the US than incentivized. The US federal government literally pays EV buyers $7500 to *not* buy a Tesla.
2) US CARB credits are occasionally worth something (far less than their nominal value), but usually worthless these days, as there's a glut on the market. The main actually profitable credits for Tesla are EU credits. Which has nothing to do with which vehicles are on the road in the US.
3) You don't need to buy CARB credits to make "planet-killing SUVs". Even if you don't earn the credits yourselves by selling clean vehicles, you can simply pay the state government the value of the credit fine. It just makes the vehicles a little bit more expensive (not even a lot more expensive). But as mentioned, even that is irrelevant, because theres a glut today, because the standards for earning credits aren't nearly stringent enough.
4) Tesla is EBIT positive even without all credits, and despite the heavy anti-incentive in the US.
5) Even if it weren't, you optimize to the system as it is. If it were the case that selling more vehicles for a lower margins was better for credits, and you took the credits away, what do you do? You sell fewer vehicles for higher margins. You don't just take the current balance and subtract the credits from it, as if nothing about how the business was run would have changed.
6) We're also playing this game of ignoring how incredibly hard it is to grow as rapidly as Tesla has, and still be profitable at the same time. It's bloody amazing that Tesla is able to do this.
Musk literally founded and runs a public transit company. It's building Las Vegas's new public transit system right now. It's IMHO such crazy purity trolling to say that if you support a mixed public transit / private cars system, or one that doesn't involve cramming as many people together in as uncomfortable of conditions as possible, that means that you want to "abolish public transit". It's sheer nonsense.
His actual quote, when asked about current types of public transit, was:
Which is why he's working on PRT [wikipedia.org] - Personal Rapid Transit. Small capsules which go where you want, when you want, and not full of a bunch of random strangers. But again, I guess because it's not miserable enough, it doesn't count as public transit, even though it's vehicles you don't own that anyone can access?
Surely in the age of COVID, the benefits of not cramming a ton of strangers together in an enclosed space should be increasingly obvious.
Re: (Score:3)
Musk literally founded and runs a public transit company. It's building Las Vegas's new public transit system right now.
To be clear, that public transit system is a tunnel which people will drive Tesla cars on.
Re:Am I missing something here? (Score:4, Informative)
His point seems to be more centered around objecting to the idea that public transit could somehow be replaced by fleets of self driving vehicles that are only (apparently) low-occupancy.
When in fact, public transit seems to me to offer the virtually perfect opportunity *FOR* autonomous vehicles, as not needing to pay salaries to drivers would enable public transit to efficiently service larger areas and with greater frequency,
The Docklands Light Railway in London has had semi-autonomous running since 1987. Trains are have the doors closed (after checking it's safe) and given a go signal by the Train Captain, and it then drives itself (at appropriate safe speed) until the next station, with the Train Captain opening the doors.
The trains can be driven manually, but are mostly left to be fully automatic. The centralised control room monitors train positions.
So, still some human intervention, but not as many staff required as having separate drivers. Mostly autonomous seems to work, with 120 million journeys per year.
Re: (Score:3)
The Skytrain in Vancouver has always been fully automated and debuted in '86. In 2016 it was the longest automated driverless system in the world, since surpassed by a couple in SE Asia. As of last year, it carried over half a million people a day.
https://en.wikipedia.org/wiki/... [wikipedia.org]
Re: (Score:2)
I think he's saying more that if all transit were low-occupancy self-driving cars, the math of them having to drive the extra distance between picking up fares would mean there would literally not be enough room on the roads anymore. (If you live in Seattle, New York City, San Francisco, Los Angeles, or Washington, D.C. this projection doesn't even seem a little absurd because there's clearly not enough room for the cars already here.)
Re: (Score:2)
I try never to read him, so thanks for doing that. If your representation is accurate then his complaint is silly. Being able to use lower-occupancy vehicles is a major feature of self-driving public transportation because heavy vehicles do basically all the road damage that isn't done by weather. Our 30' transit bus weighed over ten tons unloaded with the seats in it. All that weight is only on six tires, as it's a dually but has no tag axle. It carried 31 people (or less if it had wheelchairs aboard) whic
Re: (Score:2)
Re: (Score:2)
Yes but he's not wrong that silicon valley / capitalist elites are too blind to see stuff like that.
All indications are that Elon is brilliant, *except*, when you see how he complains about traffic in LA and doesn't realize that the biggest contributor to that is useless rush hour cubicle office culture. He only sees a car solution to car problems. Kinda makes it obvious that he's not that smart, or has a blind spot the size of Texas.
Re: (Score:3)
You can only create solutions to fixable problems. Which is easier, building cars that don't spew exhaust gasses and some tunnels to make traffic patterns 3D, or completely changing the culture of literally every business in the world with more than 40 employees?
Re: Am I missing something here? (Score:2)
Light rail is stupendously expensive. I approve of expanding it but when I needed to get to the airport one day the light rail couldn't get to the airport because there was a breakdown on the tracks. That meant no trains. So we all transferred to a bus and were eventually shuttles to the other side of the breakdown where we could ride lightrail the rest of the way.
I missed my flight and stopped taking the train to the airport. If there is a failure on a road and it's self driving you drive around it.
Ultima
Running out of space? (Score:5, Funny)
No, Musk already solved that problem by building an underground freeway. And when that freeway fills up as urban freeways inevitably do, he'll just build another under it. And another, and another. It will be tunnels all the way down!
Re: (Score:2)
You might be able to build enough tunnels to solve Southern California’s current traffic volume, but full self driving would turn Palm Springs into a suburb of LA (~125 miles). I think that is where you start to see the problem— if you marginalize public transport and at the same time encourage people to spread out more by eliminating the penalty of commute time you are going to always fight congestion. It is exactly like adding a lane to a highway; capacity is used within months of it opening
Re: (Score:2)
Re: (Score:2)
" It will be tunnels all the way down!"
Until you reach the turtle. The elephants might balk as well.
Re: (Score:2)
I don't think the tunnels are intended for the likes of us. They are intended for Musk and his peers, who would like a way to travel around in the city without having to share a room with the plebeians. Private helicopters are just too noisy, and not every mansion has a helipad.
WUT ?! (Score:2)
"Multiply the space occupied by all those AVs by the journeys people in cities need to make by the additional distances of those journeys if we need road for all those cars, and you run out of space"
Maybe whining geniuses like Cory know better than fools like Elon, who are doing nothing for this world, but dimensional analysis tells us that this sum would actually mean you run out of Space x Length.
L^3 x L = L^4
Stick to silly stories, Cory.
Re: (Score:3)
"Multiply the space occupied by all those AVs by the journeys people in cities need to make by the additional distances of those journeys if we need road for all those cars, and you run out of space"
Maybe whining geniuses like Cory know better than fools like Elon, who are doing nothing for this world, but dimensional analysis tells us that this sum would actually mean you run out of Space x Length.
L^3 x L = L^4
Stick to silly stories, Cory.
Yeah. Although it's possible that he meant the number of journeys.
Re:WUT ?! (Score:5, Informative)
His observation is basically that if you replaced a major bus route with individual cars it would be stupid.
Route 18 in london, is 9 miles long, operates 40+ busses at peak capacity, with ~150+ passengers per bus. 40 x 150 passengers = 6000 riders on the route at once at peak capacity.
Replace that with individual cars... you can only line up ~2800 Tesla 3's bumper to bumper in 9 miles, but you'd need 6000+.
And so he concludes that the idea of replacing public transit with automated cars wouldn't fit on the road; Replacing the 40+ busses on the route with the thousands of cars it would take to meet that peak load literally wouldn't fit.
Now sure you can argue that downtown commuter cars could be a smaller than a tesla 3, and that you can pack them into multple lanes, and that some people travelling together could share one, but that doesn't really counter the argument. Even if you come up with a way to pack all those extra cars into the space, it's still a very stupid plan.
Re: (Score:3)
The other argument would be that mass transit inherently combines trips on a single road, inflating the rider count on that route. If people are in individual vehicles they aren't necessarily all going the same place as everyone else, and the individual vehicles can spread out on other routes, better leveraging the entire road network rather than that one congested route.
How that affects the math is beyond me, but the only answer isn't "substitute all the bendy buses with bumper-to-bumper cars on the exact
Shamless book plug? (Score:5, Funny)
Re: (Score:3)
/. has ALWAYS loved Mr Doctorow's pronunciations, and dutifully (cheerfully) slashvertised for him forever. (shrug)
Re: Shamless book plug? (Score:3)
Except we already know Musk is a blithering idiot and almost invariably wrong about, well, everything, and a sci-fi writer is usually right about needing to sell more books.
Maybe Cory is showing a lack of imagination (Score:3)
Yes, funny thing to say about a science fiction writer but, for instance ...
Re: (Score:2)
Or from a slightly different perspective, public transit would exist in nearly its current form, just with autonomous vehicles. High-capacity vehicles with mostly predefined routes that would be extremely inexpensive to use compared to the cost of "private transit" vehicles. Maybe someday transportation can be egalitarian, but it's not likely to happen soon, even with autonomous vehicles.
Re:Maybe Cory is showing a lack of imagination (Score:4, Insightful)
there would be the option of calling on suitable vehicles.
Why does everyone conflate autonomous, self driving vehicles with shared ones? I'm not putting my self driving Bentley into a pool so some homeless bum can urinate in the back seat.
Drive near me! Braking distance! (Score:4, Insightful)
Please tell me you don't have a driver's license.
At 65 MPH, even in perfect conditions it will take your car over 300' feet to stop. In the rain, two or three times as much.
If you think that your car being 3' shorter because of no back seat means you can drive twice as close to me, doubling the number of vehicles, please, PLEASE don't drive.
In good driving conditions, for speeds between 55 and 75 mph, you should have at least four seconds between cars. That's about 380 feet. Your car being three feet shorter doesn't change that.
Following too closely is a leading cause of traffic accidents and traffic deaths. Please, don't do that.
Re: (Score:2)
Re: (Score:2)
About this big?
https://images.app.goo.gl/7f7G... [app.goo.gl]
Re: (Score:2)
It appears that, even in a city where there were plenty of on call vehicles on call within minutes,you still envisage owning your own general purpose vehicle rather than ordering a suitable vehicle for each journey you make, not ever needing to worry about parking. My own view is that the convenience and low cost of on call vehicles will overcome the desire for most people to own vehicles that they need to garage and maintain.
Autonomous vehicles on ML? (Score:3)
While autonomous vehicles do use ML for computer vision and other sensors, Iâ(TM)m not sure itâ(TM)s accurate to say they make DECISIONS with ML (that would be RL).
The decisions they make are based on path finding and obstacle avoidance rules. So the article seems to be implying something that isnâ(TM)t true.
The ML interprets input from the sensors, but any decent probabilistic AI or control system knows how to deal with faulty sensors
(Apparently Boeing didnâ(TM)t know this with the MAX but most AI companies arenâ(TM)t that dumb )
My Ideas! (Score:2, Insightful)
https://slashdot.org/comments.... [slashdot.org]
https://slashdot.org/comments.... [slashdot.org]
https://slashdot.org/submissio... [slashdot.org]
https://slashdot.org/comments.... [slashdot.org]
https://slashdot.org/comments.... [slashdot.org]
https://slashdot.org/comments.... [slashdot.org]
Science fiction author != tech/science expert (Score:3, Insightful)
Science fiction author != tech/science expert, this really should be obvious to anybody.
Doctorow adds that a key plot point in the third book in his "Little Brother" series (coming out in October) is "subverted, lethal autonomous vehicles."
Really, that's the plot your story? That sounds like the premise to a shitty B grade movie. "Cars versus People!" Give me a frick'n break.
I prefer to get my critiques on technology from people who actually know what they're talking about, thank you very much.
Re: (Score:2)
Really, that's the plot your story? That sounds like the premise to a shitty B grade movie. "Cars versus People!" Give me a frick'n break.
Actually it's "People vs People", with people in the first group hacking the software of the autonomous vehicles in order to selectively terrorize people in the second group.
Re: (Score:2)
Oh boy, that sounds like such a better premise than what I understood. Obviously people with self driving cars would then manipulate their car's programming
(because everyone obviously knows how to do that) to somehow "target" those without. What a compelling dystopian future!
You are really not making the case for a compelling story line.
Shitty B Grade Movie (Score:2)
https://www.imdb.com/title/tt0... [imdb.com]
Cory Doctorow is a scifi writer? (Score:2)
I thought he was just a podcast guy. Regardless - why do I care about his opinion on the topic of self-driving cars, as opposed to the opinions of (literally) anyone else?
Re: (Score:3)
He is an EFF advisor and wrote a few books. I am not sure I'd qualify them as science fiction as often the underlying science is almost already there.
Re: (Score:2)
Some of his writing is "meh", some of it is almost good, but none of his writing is great in my opinion.
What I don't like is that he's a gormless self-promoting blowhard who sells the amazing "insight" of the free-range hipster.
Cory Doctorow is 1/2 right (Score:3)
But for all the uses of road vehicles that remain; transport trucks, cars, buses and mini-buses etc., if cities for example, or certain major highways, actually required that all vehicles be self-driving (when that's ready) and also that the vehicles be part of a standard network where they talk to each others' driving and navigating computers, then the accident rate would definitely go down, and the efficiency of traffic flow would dramatically go up. Just one tiny example: It's super irritating when you are 6 cars back from the intersection and the light turns green and bad human reaction times and lack of situational awareness in the drivers in front of you means that it takes a long time for the car in front of you to start inching forward, and, oops, didn't make it through the intersection. Experienced drivers like taxi drivers are always darting lane to lane to avoid this. That's one minor congestion-causing and danger causing phenomenon that would be completely eliminated, because to the computer driving network, green means go for all the vehicles, in microseconds.
That was a lot of ranting... (Score:3)
...just to segue into an ad for a book. D-, see me after class
Doctorow is making a straw vehicle argument (Score:5, Insightful)
We will always have both mass transit and cars And eventually, those cars will be self-driving. Mass transit is for all those places that are already too crowded to serve with cars, automated or otherwise. Manhattan will never have car traffic only, because there physically isn't enough room to move all those people.
But transit has a fan-out problem that is just as inescapable. As soon as you get to the edge of the crowded city, the efficiency of even cheap bus networks drops off radically and you have to change from the train to a car. Even when you reach the edge of a Japanese city, you face countryside that is almost unpopulated. Suburbs and rural areas can be served only by car, and it in precisely in those uncrowded places where traffic automation will be possible first: long fast roads with few side streets, or spread-out residential areas with slow average speeds and well planned traffic flow.
The most difficult kind of automated driving will be in old, dense urban areas, where sheer geometry requires that car traffic now and always can only exist as a supplement to mass transit. Because of this, it will be where traffic automation is needed the least.
Re: (Score:2)
Tesla (Score:2)
Self driving is possible and will happen but not in Tesla unless they add two more cameras. The current 8 camera configuration on Tesla vehicles is not enough, that's obvious. The cameras are currently not placed in the best locations to see front cross-traffic in urban areas. The need to put a camera just behind each headlight. The camera in the Tesla logo is rear-facing and the camera on the B-pillar is too far back. It becomes dangerous to see traffic coming from the sides when you are at an intersection
time frame is important (Score:2)
AVs next week? No.
AVs by 2060? Sure. Think back to the computers (hardware and software) of 1980.
Anti-science science fiction (Score:3, Insightful)
Sad that most science fiction nowadays has been anti-science, exploration=bad, technology=bad. This is a trend over the recent decades to make you fear seeking out new things and trying to improve quality of life. They are telling you dragons are out there. Probably most modern writers failed or got kicked out of science? I don't know what it is.
Clarke's Three Laws (Score:3, Informative)
You'd think Mr. Doctorow would be familiar with them:
1. When a distinguished but elderly scientist states that something is possible, he is almost certainly right. When he states that something is impossible, he is very probably wrong.
2. The only way of discovering the limits of the possible is to venture a little way past them into the impossible.
3. Any sufficiently advanced technology is indistinguishable from magic.
Limited imagination (Score:2)
I don't think I'll be reading the books of someone with such a limited imagination. Yes, ML has limits today. Yes, replacing public transit in the way he outlines could be problematic. But to suggest that these and other issues will never be overcome is to ignore the history of technological innovation and to insult the imagination of every engineer.
Science fiction isn't about divining the difference between reality and fantasy, it's about creating a reality where the fantastic is real. The future is such a
Limited imagination -- that's the point (Score:2)
It's true Doctorow is making a point about self-dribving vehicles not replacing urban mass transit. But he's also making the point that self-driving isn't nearly so easy a problem as many people seem to think. Lots of executives out there are sounding a lot like sci-fi writers, imagining the wonderful world of autonomous vehicles, when in drab, boring reality self-driving vehicles are still struggling with pattern recognition, let alone inference, prediction, risk assessment, etc.
Not bullshit, but not really ready, and oversold. (Score:2)
AI driven cars *are* coming. They won't be perfect. They won't replace other forms of transit. And they won't be as soon as promised.
The thing is, even now in many circumstances (perhaps most) an automated car is a better driver than *I* am. (Well, I pulled my own license for daydreaming while driving a few decades ago.) My brother recently had a heart attack while driving and killed his wife and son as well as himself. Etc. Perfect isn't needed to accomplish better. And it also isn't needed to be "
Re: Not bullshit, but not really ready, and overso (Score:2)
Its just an infrastructure project for commercial vehicles MI is already prototyping it, roads designed for them:
https://www.businessinsider.co... [businessinsider.com]
Level 5 Autonomous cars are at least 10-20 yrs out (Score:2)
Let me guess (Score:2)
A step too far on SUVs (Score:2)
Without the tax credit program giant SUVs would still exist, as that is what USians believe they "need". What would not exist is Teslas, Chevy Bolts, Nissan Leafs, etc, and there would likely be no electric car programs at all until 2025-2030. The US automakers HATE electric vehicles with the burning fury of 1000 suns, and without the carrot and stick of the tax credits/mileage requirements there would be no incentive to start w
oh, good (Score:3)
It's so nice to see science fiction authors with no scientific training don't feel shy about expressing their unsupported opinions as scientific conclusions.
We have many problems with people not believing science and scientists. This right here is part of the problem. You should not make a career out of writing fiction and then make claims as if you're a professional expert in the real world versions of your fantasies. He has not done any actual technical or scientific work here (there's no science or math in his blog post, nor a link to any such). When we see this, we should collectively be thinking: "he has no idea what he's talking about."
He's pretending to speak from a position of authority, but what he's really doing is re-enforcing the incorrect idea that fame, snark, and jargon are suitable replacements for data and science.
"thought experiments" is a bullshit phrase (Score:2)
The WHOLE point of doing an _actual_ experiment is to VERIFY your Theory actually matches Reality.
Deductive reasoning is NOT an "though experiment" regardless of Einstein's stupidity.
I'd like a second opinion (Score:3)
Everyone in the world: "Corey Doctrow is Bullshit'
I'm sick of this faux-hipster douches holding forth on everything with his clickbait headline assertions.
Sooner or later, self-driving cars will be a real thing used by lots and lots of people. It's inevitable.
Who? (Score:3)
So, we're supposed to care about the opinion of someone with pretty much no credentials? Why is this even a discussion here?
From his Wiki page.
He quit high school,[6][verification needed] received his Ontario Academic Credit (high school diploma) from the SEED School in Toronto,[citation needed] and attended four universities without obtaining a degree.
Riding next to a serial killer (Score:3)
Re: (Score:2)
He appears to have woken up this morning, having apparently been asleep for 5 years, on the wrong side of the bed.
He offers nothing particularly insightful, nor any kind of justification for his position. But we're supposed to listen because he says it. I have doubts about self-driving cars in the near future, but I'm pretty sure we could make it work if we don't try to put all the smarts in the car, and start building roads to accomodate such cars (and only such cars). It will be a bitter pill to swallow,
Re: (Score:2)
Re: (Score:2)
And we don't even need physical rails. You could use invisible paint on a road. Or laser beams. Or magnets/sensors embedded in the roadway. Or a cable that runs overhead the roadway, like electric buses and trolleys use.
Or "virtual rails" made out of a series of GPS-encoded line segments. Put them into a ROM cartridge if you're worried about hackers changing them.
Re: (Score:2)
It's probably best not to depend on anything that can be physically manipulated by hackers, pranksters, and drunk people. Like road markings or signs. Given that there have been cases of people building illegal speed bumps outside their house when the local council ignores their complaints about people driving too fast, it seems a reasonable concern that someone might decided to stick up a fake speed limit or lay down school markings to reduce the noise - or lay diversion signs in a circle just for the fun
Re: (Score:2)
The point is to provide a permanent indicator of where cars can and can't drive
I'm a bit puzzled by this concept of where cars can't drive [toyota.co.uk].
Re: (Score:2)
The problems that self-driving cars run into are the weird edge-cases (pedestrians, weirdly painted lines, etc), and there are a lot of weird edge cases.
Re: (Score:2)
Doom and Gloom (Score:2)
The world is awash with preachers of Doom and Gloom, Fire and Brimstone.
The Baby Doomers are running amock.
There are lots of industrial control systems that can kill people if not properly maintained.
Re: (Score:3, Insightful)
The world is awash with preachers of Doom and Gloom, Fire and Brimstone.
The Baby Doomers are running amock.
There are lots of industrial control systems that can kill people if not properly maintained.
Sure, but I will never ride in a self driving car. If I'm going to die in a traffic accident it will be because of my stupidity, not because of the incompetence of some sweatshop programmer working 12 hour days 6 days a week, and 6 hours on Sundays in a run down industrial building in some emerging economy because some car company executive decided he has worked the bugs out of the outsourcing business model that led to Boeing's MCAS disaster.
Re:Doom and Gloom (Score:4, Insightful)
You forgot an equally plausible scenario: dying in a traffic accident because of someone else's stupidity. You can be the best driver in the world and still not be able to do anything about a twit rear-ending you at a stop light because they are far too busy playing with their fucking phone to pay attention to operating their motor vehicle.
Self-driving cars fix that.
Re: (Score:3)
Sure, but I will never ride in a self driving car. If I'm going to die in a traffic accident it will be because of my stupidity, not because of the incompetence of some sweatshop programmer working 12 hour days 6 days a week, and 6 hours on Sundays in a run down industrial building in some emerging economy because some car company executive decided he has worked the bugs out of the outsourcing business model that led to Boeing's MCAS disaster.
Would you feel safer in a world full of cars driven by humans with smartphones?
And standards (Score:2)
Virtually all mission critical systems are obliged to follow standards for documentation, standards for hardware, standards for specifications, standards for implementation and standards for testing.
In the short story Car Wars, every single one of these is violated.
In the Boeing software, that is again true.
In the Arianne rocket crash, implementation and testing were defective.
In the solar dust probe crash, the hardware and testing were defective (sensor mounted upside down).
Follow these procedures correctl
Re: (Score:3)
The currently most 'advanced' self-driving cars use black-box neural-net training to "learn" how to handle road situations. That means that those self-driving automobiles cannot follow explicit standards for documentation, hardware, specifications, implementation, and testing.
Driving situations, tasks, and enviro
Re: And standards (Score:4, Insightful)
"Perfect management" listens to their engineers
"Perfect management" doesn't exist.
No system should rely on any input to be "perfect".
not ship until critical flaws are properly fixed.
How do you know when all the bugs are fixed?
Answer: You don't.
The only way to be 100% sure is to never ship anything.
Re: (Score:2, Interesting)
The world is awash with preachers of Doom and Gloom, Fire and Brimstone. ...There are lots of industrial control systems that can kill people if not properly maintained.
It's not the industrial control systems that are the problem. It's the out of control economic and political control systems that are the issue. There's no actual self-restraint, no feedback, no legislative controls. The upper class has so corrupted the system of governance, and now they're completely out of control. The unrestrained greed and arrogance of the overly-affluent is going to bring this civilization crashing down soon. It's already clearly in decline, after all endless greed, complete irresponsi
Re: Self-driving cars think Cory D is BS (Score:4, Funny)
Hello there, Ameritard!
Everyone in the civilzed western world has a vast and well-working public transport system.
Amd every human is a social being, caring for his community.
Some day, you will join us civilized human people too.
Re: (Score:2)
30 miles in 30 minutes sounds a bit unrealistic. That is an average speed of 60mph across the entire route. I'm sure you can find a route where it holds true - where both your origin and destination are located directly adjacent to a freeway on-ramp, and there's no traffic. I'm not sure how that's less of a special case than living next to a bus stop. Besides, there are plenty of places with a bus stop on every block. I've yet to see anywhere manage a freeway on each block.
Re: (Score:2)
In short public transit sucks and there isn't a away to solve that problem.
Forget automated driving, there is no automation whatsoever in public transit. There is tons of opportunity for improvement. Buses are on set schedules like trains from 200 years ago. Of course that's not efficient! Demand is highly dynamic. Uber handles that with their app. Taxis suck. Mostly because they use dispatchers / century old methods. If public transit had automation, used an app to track the number of people waiting for a pickup, as well as scheduled demand, and resolves that versus their supply
Re: (Score:2)
One way to test that would be to see if there is any correlation between the "badness" of the protests, and the people leaving. For example, San Francisco hasn't had nearly as many difficulties as Seattle (because SF is used to dealing with all kinds of protests appropriately. The police have practice).
Re: (Score:2)
Only time will tell. But it's a bad time for a new technology's profitability to hinge on getting a city's public transportation funds.
Re:Cities getting less attractive-REI selling HQ (Score:2)
https://www.snewsnet.com/gear/... [snewsnet.com]
This says they aren't coming back, at least not like before.
Re: (Score:2)
So do cars with drivers.