wizard69 said: Or a deer run out infront this f the lead car! It is that time of the year.
The the reality is there are numerous emergency’s that pop up while driving that makes drafting a car on the open road very dangerous. I suspect that Apple would get significant resistance from the insurance companies.
Or a moose, or a bear, if you really want trouble. I suppose most of the people working on these projects haven't ever seen one. They aren't too common in downtown San Fran or Silicon valley from my experience living there. These people live and operate in a bubble.
Hmm except that’s just some bullshit you made up. Despite your hand-wringing, Apple is staffed with incredibly smart, thoughtful people of diverse backgrounds with offices all over the country and world. They aren’t just sitting in a circle in Cupertino, and it’s incredibly naive and outright stupid to suggest they are.
Most Moose accidents (there was one here a week or so ago - outskirts of an urban area with a larger population than S.F.) happen at night or in other bad visibility conditions.
So far, at least the meager start of AI vehicles has shown them to be far, far less safe than the average human driver. AI cars have only driven a fraction the number of miles of the amount of human driven miles for a fatality to occur, and we already have at least 2 deaths I'm aware of that were totally the car's fault. And, under ideal conditions, at that.
Meager start? Humans are no better drivers now than they were in the early days of automobiles, over 100 years ago. Only the cars have improved. In the last 16 years, autonomous vehicles have gone from failing to navigate even a few miles of open desert to driving on city streets.
It's human drivers who got the meager start and will have the meager end. Humans are marvelous creatures, and it'll be quite some time before AI in general gives us serious competition, but driving a car uses only a small subset of a human's mental capacity, and is seriously constrained by sensory input. AI in cars can't help but eventually surpass us. The evolutionary trajectory is just too steep.
When I think of self-driving vehicles I think of electric vehicles, a peloton probably results in minimal cost savings.
To show the problem of a Peloton, I’ll leave you this image:
There’s a peloton of a dozen cars in the slow lane. I’m in the middle lane, and I realize my off ramp is coming up. I have a huge wall of vehicles to my right. I turn on my turn signal. No one notices because a computer is driving, and no ones eyes are on the road. What do I do? Do I accelerate or brake aggressively to attempt to get around the wall of vehicles... BOOM I didn’t make it.
Until all vehicles are self-driving the peloton idea is a disaster waiting to happen. Somehow I don’t think we’ll have self driving motorcycles... It would kind of defeat the purpose.
Or maybe you could just accept that you were not paying attention while you were driving, and take the next exit instead of blaming a “wall of cars” for blocking your entitled path?
what would you do now? Just honk, cut people off and give them the finger because you were obviously more important?
bsimpsen said: If you've ever studied cognitive psychology, you'd know that humans behind the wheel are not doing a lot of thinking. There just isn't time for it. So, it's not necessary for computers to "think", just to drive better than we do.
Your 1 in 100million miles argument ignores the fact that no single human has ever driven anywhere near that many miles. Meanwhile, by sharing their collective experience, ALL autonomous cars could have the ability to pull from the aggregate. That's an advantage no human can hope to match.
Your false assumptions about how autonomous vehicles work makes it hard for me to share your concerns.
True, that we're more using subconscious most of the time, if I understand correctly. But, when something happens to get our 'attention' we're back in thinking mode again. Depending on how quickly it all plays out, I suppose, determines how much our thinking (and experience) play into the outcome/input. It also depends on the driver, I suppose. Better drivers are more attentive more of the time.
The 1 in a 100 million miles is just a statistic to compare how 'bad' human drivers are. The point is that the collective of AI cars haven't yet put on that many miles for a comparison in their driving safety with humans. So, saying AI cars will be safer is just a guess at this point, at least statistically.
Yes, I agree that we're talking about the aggregate with AI cars (well, at least in terms of how much the various companies working in it share the data.... will they do that?). That still doesn't mean they will make their own decisions when a new situation is encountered, just that the chances one of them encountered it in the past keep going up with the more information collected.
The problem here, though, is what is the size of the pool of un-encountered situations? For example, lets say AI cars collectively build up a 'knowledge-base' of 1 billion data points, but the real world is 1 x 10^9999 potential variants the AI will face? A human doesn't need to face each particular variant in order to have a reasonably good reaction (assuming they are paying attention). While the AI car wins in terms of paying attention, it loses on knowing what to with all that data it is collecting.
Why has Tesla not spoken of a Morphie Juice Pack arrangement for its vehicles? Even a small motorcycle style trailer with a battery pack that could double or triple the miles or even recharge itself some how.
Yeah, I've wondered that too. That would make so much sense for EVs, especially given the argument over taking trips and such. If you had enough distance (easily) for the typical commute, and then could just hook on a small trailer you would rent when going on a trip, and maybe just exchange it at certain stations along the way. What's not to like?
So Peloton is a self-driving vehicle systems. I take that is what Apple gonna sell directly to customers instead of car manufacturer. Plop the device in, and it turns your 'normal average standard' car into 'self-driving' car. Future.
No, I highly doubt that. Probably just some technology they are developing to sell to car makers.
StrangeDays said: Hmm except that’s just some bullshit you made up. Despite your hand-wringing, Apple is staffed with incredibly smart, thoughtful people of diverse backgrounds with offices all over the country and world. They aren’t just sitting in a circle in Cupertino, and it’s incredibly naive and outright stupid to suggest they are.
It's an exaggeration, but there is a bit of truth to it. I lived there for about 7 years and you kind of start forgetting how it is elsewhere... and especially when it comes to worldview or typical lifestyle, etc. it has a pretty big influence on the people there longer term.
Diversity is about a lot more than melanin count, where you geographically came from, or the food you eat. In some ways, the SF area is one of the most non-diverse places I've ever been.
mknelson said: Most Moose accidents (there was one here a week or so ago - outskirts of an urban area with a larger population than S.F.) happen at night or in other bad visibility conditions.
Lidar/IR cameras pretty much counter that.
Oh yeah, I'm quite excited about the possibility of getting systems in cars that might help to point out such things to me as a driver. I just don't want it doing the driving. And, w/o all the right systems, the AI might have worse luck in certain conditions. (i.e.: if it can't 'notice' a tractor trailer sitting across it's path, why a moose?)
bsimpsen said: Meager start? Humans are no better drivers now than they were in the early days of automobiles, over 100 years ago. Only the cars have improved. In the last 16 years, autonomous vehicles have gone from failing to navigate even a few miles of open desert to driving on city streets.
It's human drivers who got the meager start and will have the meager end. Humans are marvelous creatures, and it'll be quite some time before AI in general gives us serious competition, but driving a car uses only a small subset of a human's mental capacity, and is seriously constrained by sensory input. AI in cars can't help but eventually surpass us. The evolutionary trajectory is just too steep.
Well, I'm not sure I'd agree that humans are no better drivers now (at least the ones paying attention) than they were 100 years ago, but I get your point. And, yes, we have improved car technology and safety technology a lot... though we drive faster and more too.
Where I disagree, is that AI cars have made that improvement. Humans have made those improvements to AI and sensors.** And, that's kind of my point.... AI isn't like a human in terms of actual learning and improving. There is a qualitative difference between what a human and an AI is doing when we talk about 'driving.'
The sensors/computer/AI are going to be better at certain aspects, while the human will be better at others. In fact, there are certain things that either are absolutely incapable of doing. Pluses and minuses on both sides.
** If you put that original AI car out in that desert for another eon, it would still be out there driving quite badly. It wouldn't now be driving in the city. Humans did that, not AI. AI, in that sense, is sci-fi. 'Evolution' is even more of a joke in that sense than it is in terms of humans.
sflocal said: Or maybe you could just accept that you were not paying attention while you were driving, and take the next exit instead of blaming a “wall of cars” for blocking your entitled path?
what would you do now? Just honk, cut people off and give them the finger because you were obviously more important?
To be fair, there currently wouldn't be such a thing... because they'd all be taken out rather quickly and win the Darwin award.
And... in that, I guess is where I give hope to AI in becoming better than humans (on average)... is that human drivers seem to getting ever more stupid these days. I get nervous driving around the city often in terms of how close people are following me at speed (or hundreds of other really dumb things). I doubt they realize, for example, that if we both hit the brakes, their car will take like 10ft more distance to stop then mine, which will put them inside my trunk.
While AI might be blissfully unaware, it has the advantage of not being horrifically stupid or careless. But, the solution should be just to get more of the idiots off the road (if safety were the actual concern, which it is not).
Comments
Lidar/IR cameras pretty much counter that.
It's human drivers who got the meager start and will have the meager end. Humans are marvelous creatures, and it'll be quite some time before AI in general gives us serious competition, but driving a car uses only a small subset of a human's mental capacity, and is seriously constrained by sensory input. AI in cars can't help but eventually surpass us. The evolutionary trajectory is just too steep.
what would you do now? Just honk, cut people off and give them the finger because you were obviously more important?
The 1 in a 100 million miles is just a statistic to compare how 'bad' human drivers are. The point is that the collective of AI cars haven't yet put on that many miles for a comparison in their driving safety with humans. So, saying AI cars will be safer is just a guess at this point, at least statistically.
Yes, I agree that we're talking about the aggregate with AI cars (well, at least in terms of how much the various companies working in it share the data.... will they do that?). That still doesn't mean they will make their own decisions when a new situation is encountered, just that the chances one of them encountered it in the past keep going up with the more information collected.
The problem here, though, is what is the size of the pool of un-encountered situations? For example, lets say AI cars collectively build up a 'knowledge-base' of 1 billion data points, but the real world is 1 x 10^9999 potential variants the AI will face? A human doesn't need to face each particular variant in order to have a reasonably good reaction (assuming they are paying attention). While the AI car wins in terms of paying attention, it loses on knowing what to with all that data it is collecting.
Yeah, I've wondered that too. That would make so much sense for EVs, especially given the argument over taking trips and such. If you had enough distance (easily) for the typical commute, and then could just hook on a small trailer you would rent when going on a trip, and maybe just exchange it at certain stations along the way. What's not to like?
No, I highly doubt that. Probably just some technology they are developing to sell to car makers.
It's an exaggeration, but there is a bit of truth to it. I lived there for about 7 years and you kind of start forgetting how it is elsewhere... and especially when it comes to worldview or typical lifestyle, etc. it has a pretty big influence on the people there longer term.
Diversity is about a lot more than melanin count, where you geographically came from, or the food you eat. In some ways, the SF area is one of the most non-diverse places I've ever been.
Oh yeah, I'm quite excited about the possibility of getting systems in cars that might help to point out such things to me as a driver. I just don't want it doing the driving. And, w/o all the right systems, the AI might have worse luck in certain conditions. (i.e.: if it can't 'notice' a tractor trailer sitting across it's path, why a moose?)
Well, I'm not sure I'd agree that humans are no better drivers now (at least the ones paying attention) than they were 100 years ago, but I get your point. And, yes, we have improved car technology and safety technology a lot... though we drive faster and more too.
Where I disagree, is that AI cars have made that improvement. Humans have made those improvements to AI and sensors.** And, that's kind of my point.... AI isn't like a human in terms of actual learning and improving. There is a qualitative difference between what a human and an AI is doing when we talk about 'driving.'
The sensors/computer/AI are going to be better at certain aspects, while the human will be better at others. In fact, there are certain things that either are absolutely incapable of doing. Pluses and minuses on both sides.
** If you put that original AI car out in that desert for another eon, it would still be out there driving quite badly. It wouldn't now be driving in the city. Humans did that, not AI. AI, in that sense, is sci-fi. 'Evolution' is even more of a joke in that sense than it is in terms of humans.
To be fair, there currently wouldn't be such a thing... because they'd all be taken out rather quickly and win the Darwin award.
And... in that, I guess is where I give hope to AI in becoming better than humans (on average)... is that human drivers seem to getting ever more stupid these days. I get nervous driving around the city often in terms of how close people are following me at speed (or hundreds of other really dumb things). I doubt they realize, for example, that if we both hit the brakes, their car will take like 10ft more distance to stop then mine, which will put them inside my trunk.
While AI might be blissfully unaware, it has the advantage of not being horrifically stupid or careless. But, the solution should be just to get more of the idiots off the road (if safety were the actual concern, which it is not).