7 ms·

I was just about to comment on the same section. This kind of poorly thought out assertion throws a lot of shade on the rest of the article.

gambiting 11y ago
I don't think it does. Sure, a car would know that a stop sign was there the last time it drove through that intersection, but you can always order the car to go someplace new. And I don't believe the dream that "oh, automatic cars will just have 100% accurate street data for every road in the civilized world". First of all because we don't have this data, and even if we did, I can't imagine a car hauling several terabytes of data with it(and how would you even update it, given that our internet connections seem to be getting worse, not better every year, with new data and speed caps being introduced?). It has to be able to decide autonomously, and since it has to be programmed to do something, it's a really hard problem. A human can just "use best judgement", but a computer can't - it either follows the rules or it doesn't. And roads are not a computer simulation - they are dynamic, they change, they require the sort of input that computers are horrible at, visual recognition is dreadful, our best, cream of the top software can't tell a difference between a zebra and a sofa in a zebra print, and yet most people just wave their hand at the problems facing automatic cars saying "oh, we will just use a machine learning algorithm to watch how humans drive and then follow that". Sure. But it doesn't magically solve the problems of responsibility where a computer does something it shouldn't have.
- stormbrew 11y ago
  This is not the issue with that statement at all, and for the record I'm skeptical of cars sharing data to augment what they see (because this leaves them open to attacks that are unique to self-driving cars). The problem with the statement is that it, as the GP post says, ignores the fact that a human driving a car is no less likely to barrel through an intersection with a missing stop sign. There are a few other low level problems with this scenario as well, mostly hinging on the weird idea that self-driving cars are just going to barrel into an intersection without a stop sign as if there are no other cars going through it. This is why they have radar/lidar/whatever else.
- JoeAltmaier 11y ago
  I don't see any issue whatsoever with a car carrying around terabytes of data. That's old-think. It will never again be difficult to put startling amounts of data into devices. And the only data that needs to be kept updated, is the local environment. The car is in one place at a time. So bandwidth should not be the issue it would seem.
  - gambiting 11y ago
    You are talking about a vehicle that would have to carry dozens if not hundreds of terabytes of data. The boot would literally have to be filled with hard drives. If you are in the US you should be able to tell it to drive from one coast to the other, or if you are in Europe the car should be able to drive from Warsaw, Poland, to Aberdeen, Scotland. And what do you mean only local data needs to be updated? Why? Updating a "dumb" TomTom map takes hundreds of megabytes a month just for one country, and we are talking about much more complex data, not just a 2D map.
    - JoeAltmaier 11y ago
      It needs one large read-only copy of data, plus one hard drive with updates. Those updates are limited to say a 1000 mile range (most people don't travel 100 miles from home in a month; in any case most people travel the same routes)
      - gambiting 11y ago
        Sure, but if we want automatic cars to eventually replace manual cars we need to make sure they cover 100% use cases, not just most of them. Some people drive 10 miles to work and back, some people do 1000 miles a week driving through multiple countries. You need to deal with both, because nothing is stopping me from hopping in my car right this second and driving across the continent if I wish to do so. I wouldn't want an automatic car to hinder that freedom either.
        macjohnmcc 11y ago
- utexaspunk 11y ago
  Obviously any driverless car is not going to rely on street markings and signage alone. It's going to have sensors in all directions and see that a vehicle is headed toward the intersection at a speed which suggests it will not be able to stop in time to avoid a collision. That doesn't require some magical machine learning algorithm, and it will probably watch for this better than humans do. We tend to assume everyone else is following the rules and often don't notice when someone is about to run the stop sign and hit us.
- mikeash 11y ago
  Uncontrolled intersections are a thing. Driverless cars will have to handle them as a matter of course, simply to be competent on the roads in the first place. They won't be able to simply assume that "no stop sign" means "no stop." This scenario is just an intersection which became uncontrolled through vandalism, rather than being built that way.
  - galdosdi 11y ago
    Yes! This. Some commenters have never heard of driving defensively it seems. Do they really just barrel through uncontrolled intersections if they don't see a stop sign? For that matter, do they really just unthinkingly barrel through greens, always? If so, it probably means you live outside of the city, where you can grow accustomed to trusting that nothing unexpected will happen. In the absence of any signage, there is always an implicit "Yield." -- ie, no requirement to stop, but you should look for conflicting traffic, and preemptively slow down enough so you could stop if there turns out to be conflicting traffic. Heck, in some places even a green light is worth defensively treating as a Yield. In Manhattan, when driving, I barrel through green lights when crossing streets while going up avenues (in center lanes), but treat it more like a Yield, slowing to a quickly-stoppable speed, when going crosstown on streets across an avenue intersection. So what if you have a green light? In practice all three road user types, pedestrians, cyclists, and motorists run reds all the time. You're supposed to drive defensively, and a good driverless car should too. Anyway, I think driverless cars have an implicit advantage here in that their reaction times will be substantially lower. Hopefully they can use statistical evidence to figure out rules like the above ("Treat greens on 3+ lane roads as "No Stopping" sign, but on smaller roads, treat them as yields")
- harshalizee 11y ago
  I don't think you seem to understand how self driving cars work, or even how computers work. Firstly, the car itself need not have traversed the same road for it to have full information. It can be gathered from collective data from all the self driving cars. That means even if one car went through that intersection, it has some relevant data on it. Secondly, you're handwaving that humans are better with the "use best judgement", which is pretty much unquantifiable. Not to mention humans have "used best judgement" and still managed to crash into things. How are you sure that the algorithmic path taken by the onboard computer isn't far superior to best judgement? We already know that computers outclass us in terms of reaction times, why not this? Thirdly, you don't need to store terabytes of data to do anything in the same way you don't need to download the entire database of HN to browse the front page. Localized data would be sufficient. Even if everything is offline, the car would most likely just revert to its learning mode .. which is how most of the Google cars that are on the streets currently started off. Lastly, you're right we aren't 100% there when it comes to distinguishing zebras and couches. But for self driving cars, we don't need that. What we need is object detection, its velocity relative to us, and data on where its heading. All of which we have become pretty good at.
  - gambiting 11y ago
    Regarding your last point - we sort of do need accurate detection. What if there is a balloon on the road? I wouldn't swerve to avoid hitting a balloon, but an automatic car is going to see a round, rather large object. It might as well be a 100kg rock, hitting which would be disastrous. What is the car going to do? Stop and tell you to investigate? How about piles of leaves, cardboard boxes, large soda bottles. If you live anywhere with wind you are going to get those on the road. As for the "best judgement" part - if a human is about to get into an accident, they are going to do what feels best. It's not necessarily the best option, but no court would ever tell you "our simulation tells us that if you reacted 0.1s earlier you could have turned the wheels 30 degrees and hit a car with two occupants instead of four, who had better survival chances". A computer is going to make an algorithmic choice, and that's what the article was about - who makes the choice as to how the car should react? The owner? The manufacturer? The police? The insurer? Should it prioritize its owner life at the potential cost of other lives? Or minimise the damage to protect the interests of its insurers? We don't think about cars in those categories because we don't(and I would say we can't) make those choices when we drive. But computers will have to be programmed in some way. A lack of stop sign is a specific type of intersection in itself(where anyone going from the right has priority but anyone coming from the left has to yield) - so the car would need to observe and learn. I'm not saying it's impossible, but it's very hard to do.
jfoster 11y ago
There's a lot of clicks & eyeballs associated with FUD articles about next-gen cars at the moment. (autonomous, electric, or both) If you want to draw in a techie audience with clickbait, there's not much better you can do at the moment than writing about how Tesla has some sort of unforeseen flaw whereby both their electric & autonomous tech is doomed to failure.

gambiting 11y ago

I don't think it does. Sure, a car would know that a stop sign was there the last time it drove through that intersection, but you can always order the car to go someplace new. And I don't believe the dream that "oh, automatic cars will just have 100% accurate street data for every road in the civilized world". First of all because we don't have this data, and even if we did, I can't imagine a car hauling several terabytes of data with it(and how would you even update it, given that our internet connections seem to be getting worse, not better every year, with new data and speed caps being introduced?). It has to be able to decide autonomously, and since it has to be programmed to do something, it's a really hard problem. A human can just "use best judgement", but a computer can't - it either follows the rules or it doesn't. And roads are not a computer simulation - they are dynamic, they change, they require the sort of input that computers are horrible at, visual recognition is dreadful, our best, cream of the top software can't tell a difference between a zebra and a sofa in a zebra print, and yet most people just wave their hand at the problems facing automatic cars saying "oh, we will just use a machine learning algorithm to watch how humans drive and then follow that". Sure. But it doesn't magically solve the problems of responsibility where a computer does something it shouldn't have.

stormbrew 11y ago
This is not the issue with that statement at all, and for the record I'm skeptical of cars sharing data to augment what they see (because this leaves them open to attacks that are unique to self-driving cars). The problem with the statement is that it, as the GP post says, ignores the fact that a human driving a car is no less likely to barrel through an intersection with a missing stop sign. There are a few other low level problems with this scenario as well, mostly hinging on the weird idea that self-driving cars are just going to barrel into an intersection without a stop sign as if there are no other cars going through it. This is why they have radar/lidar/whatever else.
JoeAltmaier 11y ago
I don't see any issue whatsoever with a car carrying around terabytes of data. That's old-think. It will never again be difficult to put startling amounts of data into devices. And the only data that needs to be kept updated, is the local environment. The car is in one place at a time. So bandwidth should not be the issue it would seem.
- gambiting 11y ago
  You are talking about a vehicle that would have to carry dozens if not hundreds of terabytes of data. The boot would literally have to be filled with hard drives. If you are in the US you should be able to tell it to drive from one coast to the other, or if you are in Europe the car should be able to drive from Warsaw, Poland, to Aberdeen, Scotland. And what do you mean only local data needs to be updated? Why? Updating a "dumb" TomTom map takes hundreds of megabytes a month just for one country, and we are talking about much more complex data, not just a 2D map.
  - JoeAltmaier 11y ago
    It needs one large read-only copy of data, plus one hard drive with updates. Those updates are limited to say a 1000 mile range (most people don't travel 100 miles from home in a month; in any case most people travel the same routes)
    - gambiting 11y ago
      Sure, but if we want automatic cars to eventually replace manual cars we need to make sure they cover 100% use cases, not just most of them. Some people drive 10 miles to work and back, some people do 1000 miles a week driving through multiple countries. You need to deal with both, because nothing is stopping me from hopping in my car right this second and driving across the continent if I wish to do so. I wouldn't want an automatic car to hinder that freedom either.
      - macjohnmcc 11y ago
utexaspunk 11y ago
Obviously any driverless car is not going to rely on street markings and signage alone. It's going to have sensors in all directions and see that a vehicle is headed toward the intersection at a speed which suggests it will not be able to stop in time to avoid a collision. That doesn't require some magical machine learning algorithm, and it will probably watch for this better than humans do. We tend to assume everyone else is following the rules and often don't notice when someone is about to run the stop sign and hit us.
mikeash 11y ago
Uncontrolled intersections are a thing. Driverless cars will have to handle them as a matter of course, simply to be competent on the roads in the first place. They won't be able to simply assume that "no stop sign" means "no stop." This scenario is just an intersection which became uncontrolled through vandalism, rather than being built that way.
- galdosdi 11y ago
  Yes! This. Some commenters have never heard of driving defensively it seems. Do they really just barrel through uncontrolled intersections if they don't see a stop sign? For that matter, do they really just unthinkingly barrel through greens, always? If so, it probably means you live outside of the city, where you can grow accustomed to trusting that nothing unexpected will happen. In the absence of any signage, there is always an implicit "Yield." -- ie, no requirement to stop, but you should look for conflicting traffic, and preemptively slow down enough so you could stop if there turns out to be conflicting traffic. Heck, in some places even a green light is worth defensively treating as a Yield. In Manhattan, when driving, I barrel through green lights when crossing streets while going up avenues (in center lanes), but treat it more like a Yield, slowing to a quickly-stoppable speed, when going crosstown on streets across an avenue intersection. So what if you have a green light? In practice all three road user types, pedestrians, cyclists, and motorists run reds all the time. You're supposed to drive defensively, and a good driverless car should too. Anyway, I think driverless cars have an implicit advantage here in that their reaction times will be substantially lower. Hopefully they can use statistical evidence to figure out rules like the above ("Treat greens on 3+ lane roads as "No Stopping" sign, but on smaller roads, treat them as yields")
harshalizee 11y ago
I don't think you seem to understand how self driving cars work, or even how computers work. Firstly, the car itself need not have traversed the same road for it to have full information. It can be gathered from collective data from all the self driving cars. That means even if one car went through that intersection, it has some relevant data on it. Secondly, you're handwaving that humans are better with the "use best judgement", which is pretty much unquantifiable. Not to mention humans have "used best judgement" and still managed to crash into things. How are you sure that the algorithmic path taken by the onboard computer isn't far superior to best judgement? We already know that computers outclass us in terms of reaction times, why not this? Thirdly, you don't need to store terabytes of data to do anything in the same way you don't need to download the entire database of HN to browse the front page. Localized data would be sufficient. Even if everything is offline, the car would most likely just revert to its learning mode .. which is how most of the Google cars that are on the streets currently started off. Lastly, you're right we aren't 100% there when it comes to distinguishing zebras and couches. But for self driving cars, we don't need that. What we need is object detection, its velocity relative to us, and data on where its heading. All of which we have become pretty good at.
- gambiting 11y ago
  Regarding your last point - we sort of do need accurate detection. What if there is a balloon on the road? I wouldn't swerve to avoid hitting a balloon, but an automatic car is going to see a round, rather large object. It might as well be a 100kg rock, hitting which would be disastrous. What is the car going to do? Stop and tell you to investigate? How about piles of leaves, cardboard boxes, large soda bottles. If you live anywhere with wind you are going to get those on the road. As for the "best judgement" part - if a human is about to get into an accident, they are going to do what feels best. It's not necessarily the best option, but no court would ever tell you "our simulation tells us that if you reacted 0.1s earlier you could have turned the wheels 30 degrees and hit a car with two occupants instead of four, who had better survival chances". A computer is going to make an algorithmic choice, and that's what the article was about - who makes the choice as to how the car should react? The owner? The manufacturer? The police? The insurer? Should it prioritize its owner life at the potential cost of other lives? Or minimise the damage to protect the interests of its insurers? We don't think about cars in those categories because we don't(and I would say we can't) make those choices when we drive. But computers will have to be programmed in some way. A lack of stop sign is a specific type of intersection in itself(where anyone going from the right has priority but anyone coming from the left has to yield) - so the car would need to observe and learn. I'm not saying it's impossible, but it's very hard to do.

jfoster 11y ago

There's a lot of clicks & eyeballs associated with FUD articles about next-gen cars at the moment. (autonomous, electric, or both) If you want to draw in a techie audience with clickbait, there's not much better you can do at the moment than writing about how Tesla has some sort of unforeseen flaw whereby both their electric & autonomous tech is doomed to failure.