If it is a choice between sharing an automated car or having a manual one in my driveway, I'll keep my manual car thanks. My wife and I have the stuff we want in our car, my kids have the stuff they want, it's our space. Not interested in taking someone else's car wherever I go.
Unless you're going to buy all road texters an automated car, they are not likely to see enough of a problem to warrant the possibly double cost of a vehicle. The economic factor is the thing that no one comprehends. People will mostly buy manual cars because they can afford them.
No, I was talking about any situation where Autopilot must relinquish control... including safety emergencies. Not sure why anyone would make a distinction between one or the other. Point being, not all of them can be done over a time period of minutes.
I'm anticipating that all these automated car makers will start to complain about steep regulations, such as having to avoid children and animals running into the road and having to keep to a common reasonable driving speed.
So then you admit putting AI on the roads today amounts to vehicular manslaughter then. Thanks for being honest. The question is whether automation will be able to survive the lawsuits, and I hope any family that loses a member will bring the fires of hell down on the company responsible for releasing that particular AI. Even more importantly, the execs of that company should have to meet with the family and explain to them why their solution was worth their son's/daughter's/father's/mother's life. Relieving people of the burden of driving doesn't quite cut it with me, because that really is a first world problem. This all just seems to be about luxury and convenience.
There was already a discussion about this. Even most people on slashdot admitted that they would not be completely productive in a perfect self driving car. The goal will always be to get from point A to point B in the fastest amount of time. Also, going unnecessarily slow is frustrating to other drivers, and might amount to a violation of impeding the flow of traffic.
it would take action to halt or swerve the moment the ball appeared
But what if that is the wrong thing to do? Currently AI cannot tell the difference between a puffed out plastic shopping bag and a ball. You don't want the car to swerve or stop for a shopping bag, since humans wouldn't.
To see to the other side of a parked car, you need to practically lie on the street. In order for the sensor to see along the road under a car, it would have to be placed below the car that is driving. So unless some sensors are going to extend downward once the car start driving, somehow able to survive bumps as the shocks of the vehicle compress, and look directly across, they will not be able to see a child coming through a row of parked cars. Unless of course they have x-ray vision that can see through metal. Any other placement will be looking at the parked car at a downward angle and will therefore be obscured. Also, it might not be a car. It might be a fence, or a community mailbox, or some other fixed object.
I've read some interesting things about bittorrent filesystems, sounds like it's the kind of thing you're looking for. Your files get encrypted and broken up into the torrentsphere. I haven't tried them personally, but some people seem to like them.
It sometimes amazes me that the very same people falling all over themselves to get an electric car, none of which are actually better for the environment at the moment anyway, will also be consuming every new electronic device known to man. Using electronics until they actually break is a valid way to help the environment as well. Also, I don't wish to be one of those people who complain about the decline of the middle class yet always seem to have the newest things. Using electronics until they break is also financially responsible, and I find it sad that Apple does not choose to support these values.
I'm looking to migrate to a small cloud server and AWS seemed very expensive to me. I was surprised that such a large company could not beat a smaller one suck as Linode on price. I have no allegiances, so Oracle is either going to cheaper and more useful, or more expensive and less useful. Most people will base their opinions solely on that. So the question is, has Larry talked price yet?
I disagree. A human is going to see a ball in the road and look for a child chasing the ball. It will be instinctual. Some other things a human might notice would be:
- A child in a yard beside the road with a baseball mitt staring at the road (at another child going for a ball)
- A yard with toys strewn about and no child (where is the child?)
- A group of children calling to (a child between the cars)
- Passing a school with children playing
These are the kinds of cues that a human picks up on very easily. I live on a block where there are a lot of kids that play, I always look for their parents or signs that they are playing such as open garage doors before I even get in my vehicle. These are cues that AI will have to pick up on in order to be at least as good as a human.
The thing that I agree with you on is that a human is less likely to slow down, but not for the reason you are suggesting. A human will drive faster in such situations because they can intuit the danger, and thus drive a more reasonable speed. They may drive slower past a row of parked cars near a park or a school, slower still if there are other children playing in that park or school yard. AI will have to have this intuition to be safe. If automation is to coexist with humans it will need to exhibit this same behavior or it will cause nothing but traffic problems.
I'm not saying that AI will never be better than a human in some situations, but that isn't the goal. The goal is for AI not to kill or injure people in every situation possible where a human would not. This means that it must have no weaknesses or scenarios where it does worse than a human, ever, and we are no where near that yet.
Lol.. So you think people will use self driving cars if they slow down every time they pass a parked car? and I repeat... What stopping distance is adequate when passing each and every parked car? Lets say the car picks a 100% safe stopping distance.. two feet. That's 7 miles per hour on ice. So now you are saying that self driving cars will be driving 7 miles per hour past each and every parked vehicle in case a child runs out. That's going to be interesting, especially when the manual cars lining up behind the automated car don't expect it.
Is it a stupid scenario? You propose that children never run out from between parked cars? That's funny because I was told not to do it at least fifty times as a child.
The question is, how MUCH does the car slow down as a result of detecting the road is slippery... let's look at it this way, what stopping distance is the car to target? It won't be a stopping distance of two feet, because then it will be driving at walking speed everywhere. Stopping distance for an average vehicle on ice at 40mph is 138 feet. So lets say AI is better than a human and call it 120 feet. A person seeing a ball can easily see it 2 seconds before the kid runs out, and in two seconds at 40mph you travel that stopping distance. So yes it will be very important that the car determines that a child may be running out.
There are always things on the road.. Cars will have to avoid some and slow down for others in order to be workable. You don't want a car to slow down for a shopping bag or a leaf blowing across the road, yet if it is a small gosling which can be the same color and size as a leaf, you do want the car to slow down.
You just aren't opening your mind to all the subtitles in the situations that occur in the real world, if it can physically happen then it will so I am really not grasping at straws here even though it may seem to be a tailored scenerio. A person would be stupid to bet their life that current AI won't cause a death or injury at some point so why bet the lives of others?
You haven't answered how this would work in terms of the scenario I presented. Road is slippery, ball bounces out from between two cars, child runs out after it. We can make the assumption there will be no 'street level' scanner that will see the child's feet under the cars so it doesn't see the child. How will it know how to stop? Will it stop and wait 15 seconds for every single thing on the road? If you don't explain the mechanics then no point in arguing that it can happen.
Another thing, the car's reaction time doesn't mean anything if the road is icy or slippery.. So will an automated car drive 10 mph everywhere if the road is slippery and a child may run out into the road and block traffic behind it? Or will it actually be able to understand the object and anticipate the child to give more stopping time? These are very common situations in the real world; I'm sure it happens somewhere a hundred times a day.
So tell me the number of sufficient cases you need to determine if an object is any kind of child's toy or not any kind of childs toy? If you don't know the amount of trials you need to do to cover any object in the world, then how do you know if you're going to be killing a child or not the first time it happens?
Slashdot Mirror
What will prevent people from just owning manual cars like they do today?
If it is a choice between sharing an automated car or having a manual one in my driveway, I'll keep my manual car thanks. My wife and I have the stuff we want in our car, my kids have the stuff they want, it's our space. Not interested in taking someone else's car wherever I go.
So people who can't afford food and shelter for their families are richer? Who knew??
I'm getting tired of these kinds of arguments.. Please explain how those 38,000 people would afford an automated car even if they were real.
Unless you're going to buy all road texters an automated car, they are not likely to see enough of a problem to warrant the possibly double cost of a vehicle. The economic factor is the thing that no one comprehends. People will mostly buy manual cars because they can afford them.
I'll believe it when I see it. Currently it is nowhere close.
No, I was talking about any situation where Autopilot must relinquish control... including safety emergencies. Not sure why anyone would make a distinction between one or the other. Point being, not all of them can be done over a time period of minutes.
I'm anticipating that all these automated car makers will start to complain about steep regulations, such as having to avoid children and animals running into the road and having to keep to a common reasonable driving speed.
It's not rare and unlikely. Autopilot doesn't drive from door to door.
Would you rather they be busy on their phone and unable to take over from Autopilot?
The epipen costs $120 in Canada, so there is still PLENTY of profit on that. Just not as obscene.
So then you admit putting AI on the roads today amounts to vehicular manslaughter then. Thanks for being honest. The question is whether automation will be able to survive the lawsuits, and I hope any family that loses a member will bring the fires of hell down on the company responsible for releasing that particular AI. Even more importantly, the execs of that company should have to meet with the family and explain to them why their solution was worth their son's/daughter's/father's/mother's life. Relieving people of the burden of driving doesn't quite cut it with me, because that really is a first world problem. This all just seems to be about luxury and convenience.
There was already a discussion about this. Even most people on slashdot admitted that they would not be completely productive in a perfect self driving car. The goal will always be to get from point A to point B in the fastest amount of time. Also, going unnecessarily slow is frustrating to other drivers, and might amount to a violation of impeding the flow of traffic.
it would take action to halt or swerve the moment the ball appeared
But what if that is the wrong thing to do? Currently AI cannot tell the difference between a puffed out plastic shopping bag and a ball. You don't want the car to swerve or stop for a shopping bag, since humans wouldn't.
To see to the other side of a parked car, you need to practically lie on the street. In order for the sensor to see along the road under a car, it would have to be placed below the car that is driving. So unless some sensors are going to extend downward once the car start driving, somehow able to survive bumps as the shocks of the vehicle compress, and look directly across, they will not be able to see a child coming through a row of parked cars. Unless of course they have x-ray vision that can see through metal. Any other placement will be looking at the parked car at a downward angle and will therefore be obscured. Also, it might not be a car. It might be a fence, or a community mailbox, or some other fixed object.
I've read some interesting things about bittorrent filesystems, sounds like it's the kind of thing you're looking for. Your files get encrypted and broken up into the torrentsphere. I haven't tried them personally, but some people seem to like them.
It sometimes amazes me that the very same people falling all over themselves to get an electric car, none of which are actually better for the environment at the moment anyway, will also be consuming every new electronic device known to man. Using electronics until they actually break is a valid way to help the environment as well. Also, I don't wish to be one of those people who complain about the decline of the middle class yet always seem to have the newest things. Using electronics until they break is also financially responsible, and I find it sad that Apple does not choose to support these values.
Please follow along. The situation is that there is a ball in the road giving a clue as to what will happen.
I'm looking to migrate to a small cloud server and AWS seemed very expensive to me. I was surprised that such a large company could not beat a smaller one suck as Linode on price. I have no allegiances, so Oracle is either going to cheaper and more useful, or more expensive and less useful. Most people will base their opinions solely on that. So the question is, has Larry talked price yet?
I disagree. A human is going to see a ball in the road and look for a child chasing the ball. It will be instinctual. Some other things a human might notice would be:
- A child in a yard beside the road with a baseball mitt staring at the road (at another child going for a ball)
- A yard with toys strewn about and no child (where is the child?)
- A group of children calling to (a child between the cars)
- Passing a school with children playing
These are the kinds of cues that a human picks up on very easily. I live on a block where there are a lot of kids that play, I always look for their parents or signs that they are playing such as open garage doors before I even get in my vehicle. These are cues that AI will have to pick up on in order to be at least as good as a human.
The thing that I agree with you on is that a human is less likely to slow down, but not for the reason you are suggesting. A human will drive faster in such situations because they can intuit the danger, and thus drive a more reasonable speed. They may drive slower past a row of parked cars near a park or a school, slower still if there are other children playing in that park or school yard. AI will have to have this intuition to be safe. If automation is to coexist with humans it will need to exhibit this same behavior or it will cause nothing but traffic problems.
I'm not saying that AI will never be better than a human in some situations, but that isn't the goal. The goal is for AI not to kill or injure people in every situation possible where a human would not. This means that it must have no weaknesses or scenarios where it does worse than a human, ever, and we are no where near that yet.
Lol.. So you think people will use self driving cars if they slow down every time they pass a parked car? and I repeat... What stopping distance is adequate when passing each and every parked car? Lets say the car picks a 100% safe stopping distance.. two feet. That's 7 miles per hour on ice. So now you are saying that self driving cars will be driving 7 miles per hour past each and every parked vehicle in case a child runs out. That's going to be interesting, especially when the manual cars lining up behind the automated car don't expect it.
Is it a stupid scenario? You propose that children never run out from between parked cars? That's funny because I was told not to do it at least fifty times as a child.
The question is, how MUCH does the car slow down as a result of detecting the road is slippery... let's look at it this way, what stopping distance is the car to target? It won't be a stopping distance of two feet, because then it will be driving at walking speed everywhere. Stopping distance for an average vehicle on ice at 40mph is 138 feet. So lets say AI is better than a human and call it 120 feet. A person seeing a ball can easily see it 2 seconds before the kid runs out, and in two seconds at 40mph you travel that stopping distance. So yes it will be very important that the car determines that a child may be running out.
There are always things on the road.. Cars will have to avoid some and slow down for others in order to be workable. You don't want a car to slow down for a shopping bag or a leaf blowing across the road, yet if it is a small gosling which can be the same color and size as a leaf, you do want the car to slow down.
You just aren't opening your mind to all the subtitles in the situations that occur in the real world, if it can physically happen then it will so I am really not grasping at straws here even though it may seem to be a tailored scenerio. A person would be stupid to bet their life that current AI won't cause a death or injury at some point so why bet the lives of others?
You haven't answered how this would work in terms of the scenario I presented. Road is slippery, ball bounces out from between two cars, child runs out after it. We can make the assumption there will be no 'street level' scanner that will see the child's feet under the cars so it doesn't see the child. How will it know how to stop? Will it stop and wait 15 seconds for every single thing on the road? If you don't explain the mechanics then no point in arguing that it can happen.
Another thing, the car's reaction time doesn't mean anything if the road is icy or slippery.. So will an automated car drive 10 mph everywhere if the road is slippery and a child may run out into the road and block traffic behind it? Or will it actually be able to understand the object and anticipate the child to give more stopping time? These are very common situations in the real world; I'm sure it happens somewhere a hundred times a day.
So tell me the number of sufficient cases you need to determine if an object is any kind of child's toy or not any kind of childs toy? If you don't know the amount of trials you need to do to cover any object in the world, then how do you know if you're going to be killing a child or not the first time it happens?
How will that field of view work for it if the ball is visible but the child is in the process of running out from between two parked cars?