|
fishmech posted:Humans successfully drive, actually. It's that simple, cars that fail way more often are not going to provide safety or other benefits. Automated cars don't fail way more often, idiot. Just the opposite. Even during the current beta period, they are much better than human drivers. Go look at the numbers.
|
# ? Aug 28, 2018 20:07 |
|
|
# ? Jun 11, 2024 00:41 |
|
ryonguy posted:What's probably going to gently caress self driving cars the most is the reality that they will work exponentially better if each one can communicate its vector and a limited portion of its future route to every other nearby vehicle and vice versa, but we'll have to put up with the autonomous vehicle equivalent of VHS vs Beta either until the "best" system "wins", or the govn't tells them to shut the gently caress up and pick a standard. Bunching up and other traffic issues don't happen because of "varying degrees of reflex reaction time", they happen because of the physical space demands of traffic maneuvers such as merges and lane changes.
|
# ? Aug 28, 2018 20:16 |
|
enraged_camel posted:Automated cars don't fail way more often, idiot. Just the opposite. Even during the current beta period, they are much better than human drivers. Go look at the numbers. ah cool, this is where people who dont know anything about statistics argue in favor of numbers which dont really exist which numbers, please? can you post them? can you explain them?
|
# ? Aug 28, 2018 20:17 |
|
Main Paineframe posted:Bunching up and other traffic issues don't happen because of "varying degrees of reflex reaction time", they happen because of the physical space demands of traffic maneuvers such as merges and lane changes.
|
# ? Aug 28, 2018 20:23 |
|
enraged_camel posted:Automated cars don't fail way more often, idiot. Wrong. Absolutely 0 of them can say come to where I am in Long Island at this moment and drive me over to Denver. That's a pretty big fail. mobby_6kl posted:There's obviously an upper bound on how much traffic can flow in a particular place, but a lot of the issues are absolutely caused by sub-optimal human behavior, e.g. And you can only add tiny amounts of capacity by "perfect" reflexes, because you're still saddled by rubber on asphalt (or concrete or whatever)
|
# ? Aug 28, 2018 20:27 |
|
mobby_6kl posted:There's obviously an upper bound on how much traffic can flow in a particular place, but a lot of the issues are absolutely caused by sub-optimal human behavior, e.g. this is going to continue to happen so long as you're mixing manual and automatic cars. and if we really wanted to mandate driverless vehicles we could do that right now, they're called trains
|
# ? Aug 28, 2018 20:28 |
|
luxury handset posted:this is going to continue to happen so long as you're mixing manual and automatic cars. and if we really wanted to mandate driverless vehicles we could do that right now, they're called trains Driverless trains exist and are used in dubai and a few other places but are pretty much as much an up and coming technology as driverless cars.
|
# ? Aug 28, 2018 21:00 |
|
fishmech posted:Wrong. Absolutely 0 of them can say come to where I am in Long Island at this moment and drive me over to Denver. That's a pretty big fail.
|
# ? Aug 28, 2018 21:10 |
|
StabbinHobo posted:what a spoiled little brat way of thinking Sure the car can service normal commutes that account for 99% of use cases, but what about the 1% of cases that aren't satisfied? More like dumb cars!
|
# ? Aug 28, 2018 21:22 |
|
Main Paineframe posted:Yeah, but there's a difference between "every car using their own mapping equipment to read the road on the fly" and "special mapping cars with much higher-quality sensors and human drivers have to regularly travel every single road to generate hi-res maps that get looked over and adjusted by human technicians and uploaded to all the self-driving cars". The latter approach is just covering up the flaws of their current self-driving algorithms by throwing lots of money and man-hours at manually creating self-driving "rails" for the cars to follow. I'm unclear why an unchanging road would need to be scanned and rescanned frequently if the layout of the road stayed the same but even in some fantasy worst case where they had to scan the roads every single day (for some unspecified reason where that is required somehow) stuff like newspaper delivery drives to addresses all over an area for decades and decades without that being any sort of technical impossibility. Having people drive everywhere is a thing businesses have done before. The fact the cars are automated is only going to make that more feasible. although again, the idea is you scan once to set up the 'rail' then don't scan again until something major enough changes that the car can't locally just navigate the obstacle then return to the rail.
|
# ? Aug 28, 2018 22:11 |
|
ElCondemn posted:Sure the car can service normal commutes that account for 99% of use cases i dont think this has been proven yet, the distribution of origin-destination pairs for taxis is somewhat different for commuting. taxis you would expect to be somewhat more randomly distributed across both time and space than work commutes because taxis tend to be used for more special and occasional trips and are a fairly expensive option for a regular fixed commute
|
# ? Aug 28, 2018 22:13 |
|
luxury handset posted:i dont think this has been proven yet, the distribution of origin-destination pairs for taxis is somewhat different for commuting. taxis you would expect to be somewhat more randomly distributed across both time and space than work commutes because taxis tend to be used for more special and occasional trips and are a fairly expensive option for a regular fixed commute Look at services like car2go as the example, not taxis.
|
# ? Aug 28, 2018 22:26 |
|
ElCondemn posted:Look at services like car2go as the example, not taxis. this is a different business model from self driving cars though. i mean, car rentals exist... and?
|
# ? Aug 28, 2018 22:31 |
|
mobby_6kl posted:There's obviously an upper bound on how much traffic can flow in a particular place, but a lot of the issues are absolutely caused by sub-optimal human behavior, e.g. What's causing the slowdown here isn't "human behavior", it's the fact that the first red car needed to lane-change when traffic was slightly too dense for smooth lane-changing, forcing other cars to slow down to allow it enough room. "Only one car can exist in a given place at any one time" and "a car can't pass through a space too small for it" aren't things a self-driving car will solve.
|
# ? Aug 28, 2018 22:41 |
|
Main Paineframe posted:What's causing the slowdown here isn't "human behavior", it's the fact that the first red car needed to lane-change when traffic was slightly too dense for smooth lane-changing, forcing other cars to slow down to allow it enough room. "Only one car can exist in a given place at any one time" and "a car can't pass through a space too small for it" aren't things a self-driving car will solve. The gif is from a video about the quirk in human driving that if three cars are stopped in a row humans drive by having the first car move away then the second car move away then the third car move , but all three cars can actually just move at once, but it takes a lot of mental work and attention to do that, so people generally don't and the same effect happens in any slowing down/speeding up situation, people wait for the car ahead of them to "finish" before they start and a trained human (race driver/military) or a computer can do it much better by not multiplying the delay for every additional car and just having every car move at once (or closer to at once since no one is psychic).
|
# ? Aug 28, 2018 22:49 |
|
ElCondemn posted:Sure the car can service normal commutes that account for 99% of use cases, They can't
|
# ? Aug 28, 2018 23:40 |
|
Owlofcreamcheese posted:Driverless trains exist and are used in dubai and a few other places but are pretty much as much an up and coming technology as driverless cars. Driverless trains have existed since the 60s. Shockingly, isolated rail networks are quite a bit different from the open road. StabbinHobo posted:what a spoiled little brat way of thinking Truly it is too much to expect a self driving car to be able to self drive. ElCondemn posted:Sure the car can service normal commutes that account for 99% of use cases, but what about the 1% of cases that aren't satisfied? More like dumb cars! This might shock you, but millions of people commute from Long Island. It's kind of the only reason Long Island exists. If a car can't handle it it is literally useless to 99% of drivers. Owlofcreamcheese posted:I'm unclear why an unchanging road would need to be scanned and rescanned frequently i The other 99% of the roads out there need to scanned and rescanned frequently.
|
# ? Aug 29, 2018 00:44 |
|
fishmech posted:The other 99% of the roads out there need to scanned and rescanned frequently. And? I'm saying that even if that was true somehow that the idea of a company needing to employee a fleet of people that drive around everywhere was even in the grasp of a company the size of like a city newspaper and the fact the surveys could be done by automated cars to check off the places that didn't change make it even less labor intensive than even pizza delivery or something.
|
# ? Aug 29, 2018 00:48 |
|
Owlofcreamcheese posted:And? I'm saying that even if that was true somehow that the idea of a company needing to employee a fleet of people that drive around everywhere was even in the grasp of a company the size of like a city newspaper and the fact the surveys could be done by automated cars to check off the places that didn't change make it even less labor intensive than even pizza delivery or something. You're just wording without meaning. It's very labor intensive to operate a massive country-mapping fleet that can update these maps even once yearly, let alone the busy areas where it'll need updates even more often. Remember that these maps need significantly more detail than, well, a simple human-used map where you can elide things like exact positioning of turns or lane changes by just telling the driver to take care of it. Perhaps you live in some empty rural expanse where no one actively changes anything on the roads in decades. In places people live, it's not uncommon to have very lengthy infrastructure projects ongoing for years on end and with where the lanes and exits actually are changing weekly.
|
# ? Aug 29, 2018 01:47 |
|
fishmech lives in Long Island? Man, that makes a lot of sense.
|
# ? Aug 29, 2018 02:33 |
|
fishmech posted:It's very labor intensive to operate a massive country-mapping fleet that can update these maps even once yearly, Internet says the US has 2.68 million miles of paved roads. If you had to map those once a year you'd need to get 7342 miles done a day. Even if you assumed an extremely slow pace like 50 miles a day driven by each driver that is still only 150 employees. That seems totally doable, even way more frequently than once a year. Especially if automated cars can just drive around checking off roads that remain unchanged (or aircraft surveys or passing traffic or whatever other data source saying "nah, nothing big changed"). Since it doesn't need to confirm every single slight change, just changes big enough to destroy the "rail" of the road that the car sticks to and deviates from for on the fly obstacles and traffic conditions.
|
# ? Aug 29, 2018 02:42 |
|
enraged_camel posted:fishmech lives in Long Island? Man, that makes a lot of sense. No, I don't. Owlofcreamcheese posted:Internet says the US has 2.68 million miles of paved roads. If you had to map those once a year you'd need to get 7342 miles done a day. Even if you assumed an extremely slow pace like 50 miles a day driven by each driver that is still only 150 employees. That seems totally doable, even way more frequently than once a year. Especially if automated cars can just drive around checking off roads that remain unchanged (or aircraft surveys or passing traffic or whatever other data source saying "nah, nothing big changed"). Since it doesn't need to confirm every single slight change, just changes big enough to destroy the "rail" of the road that the car sticks to and deviates from for on the fly obstacles and traffic conditions. Your math is way the gently caress off. You really don't seem to get that they do the intense mapping with high end equipment in the first place because the cars can't handle driving without it. Also if we're trying to rely on having the cars randomly mapping trips we're generating a shitload of additional traffic and wasting a bunch of fuel/electricity too. So congrats on inventing the bitcoin of cars.
|
# ? Aug 29, 2018 03:18 |
|
fishmech posted:This might shock you, but millions of people commute from Long Island. It's kind of the only reason Long Island exists. If a car can't handle it it is literally useless to 99% of drivers. LIRR exists too. a self driving shuttle will take you to and from the station for a loving decade before it can get you to denver. which is a goddamn shame because the correct policy is to loving drive long islanders into the sea with cattle prods
|
# ? Aug 29, 2018 03:18 |
|
StabbinHobo posted:lol you somehow moved the goal post from "my front door to denver co" to "long island commute". thats just loving nuts. you're a random nonsense shouting rear end in a top hat. My front door isn't on Long Island, also my car can already be driven to Denver just fine. That is the point, arbitrary travel, the thing people have cars for. If we're just going to rely on public transit, what is the point of spending trillions of dollars to fail to build a self driving car capable of driving anywhere an Edsel could?
|
# ? Aug 29, 2018 03:27 |
|
again, its just the purely self centered reasoning of a child. its not cool to you, and it can't do more than you can already do, so therefore its worthless to not just you but anyone and everyone else too. once it fails your "me me me" test its cast into the void, garbage, be gone.
|
# ? Aug 29, 2018 03:49 |
|
StabbinHobo posted:again, its just the purely self centered reasoning of a child. Sorry you got so offended by the idea that a self driving car should be able to drive itself? Look you're already free to buy broken vehicles all you want. It's a simple request - a car you're trying to sell should be able to handle suburbs, city, and the open highway. It shouldn't be something that can barely manage to make it to a train station, especially when it's trying to be positioned as replacing other cars.
|
# ? Aug 29, 2018 04:06 |
|
fishmech posted:
Post some math that makes the number employed actually unreasonable for a car company or google sized company. In what way is my math off? Even if you demand something absurd like an employee being able to only map 2000 feet of road a day that still wouldn’t even be the biggest department of employees for the companies like google or ford or something that would run it.
|
# ? Aug 29, 2018 04:09 |
|
Owlofcreamcheese posted:Post some math that makes the number employed actually unreasonable for a car company or google sized company. In what way is my math off? Even if you demand something absurd like an employee being able to only map 2000 feet of road a day that still wouldn’t even be the biggest department of employees for the companies like google or ford or something that would run it. You don't need to grab 7200 random miles a day and call it good, you need frequent updates in the populated 4% of the land mass of the country. Though you do need to still be checking all the rest as well on top of that. And keeping up all the maintenance and provision of the high quality sensors which must of course be kept clear to record usable data. You're also gleefully glossing over the work of recording all the roads the first time around, which definitely hasn't been done yet, and would need a truly vast amount of vehicles to get going.
|
# ? Aug 29, 2018 04:29 |
|
fishmech posted:You don't need to grab 7200 random miles a day and call it good, you need frequent updates in the populated 4% of the land mass of the country. Though you do need to still be checking all the rest as well on top of that. And keeping up all the maintenance and provision of the high quality sensors which must of course be kept clear to record usable data. I think you just don’t know poo poo about technology, the roads are very much being updated daily especially in the “populated 4% of the landmass”. They have this other service that does what you’re asking for albeit with different sensors, it’s called street view.
|
# ? Aug 29, 2018 04:33 |
|
ElCondemn posted:I think you just don’t know poo poo about technology, the roads are very much being updated daily especially in the “populated 4% of the landmass”. They have this other service that does what you’re asking for albeit with different sensors, it’s called street view. Street View is a great example! For instance, it rarely updates, often going years after a major building went in or even changes on the road happened before the new stuff shows up. It's also entirely missing vast amounts of roads, particularly newly built ones. Here, I'll show you a great example of its failure: https://www.google.com/maps/@42.251...!7i13312!8i6656 See this view? Captured September 2014 and still on there. It's also a road that's been completely demolished for a park project and hasn't been driveable since 2016. Even the basic street data wasn't updated until a few months after that bit of roadway was blocked off and demolished and Google Maps directions would tell you to use it when going through that section of the town despite it being impossible. You can also see that the surrounding street view imagery on the roads that do still exist didn't get taken until August of 2017, which was over a year after the road was closed off.
|
# ? Aug 29, 2018 04:43 |
|
fishmech posted:You don't need to grab 7200 random miles a day and call it good, you need frequent updates in the populated 4% of the land mass of the country. It's just extremely unclear what you feel is wrong with the math. There just isn't ultimately that many miles of road in the US. Texas is the state with the most miles of road and it has 303,176. Any way you chop that up just doesn't end up being that big of a number. a hundred employees could cover that in a year crawling on their hands and knees with the lidar riding on their back like a horse. There is really no way you can play with the numbers where you end up with an unreasonably high number of jobs for a large company to fill. The world has plenty of road surface but not THAT much in the grand scheme of things. New york city only has 6024 miles of road, even if you make outlandish claims that it would need to be scanned every 30 minutes incase roads suddenly changed places that still ends up not all that shockingly large a number. There just isn't really a set of requirements you can think up that make the ultimately pretty small mileage of roads that exist require a workforce that is all that high improbable massive. Despite any sort of ultra slow, ultra frequent scanning not being a requirement anything has ever actually hinted exists in any way.
|
# ? Aug 29, 2018 05:51 |
|
Owlofcreamcheese posted:It's just extremely unclear what you feel is wrong with the math. There just isn't ultimately that many miles of road in the US. Your belief that it indicates a minor effort, and also your belief it would be sufficient. You also seem to miss that you can't just expect to cover hundreds of needed miles of say city streets per car per day.
|
# ? Aug 29, 2018 05:59 |
|
fishmech posted:You also seem to miss that you can't just expect to cover hundreds of needed miles of say city streets per car per day. Okay? Even if you go to some absurd extreme where you need a separate employee for every single mile in new york city and you need to map the city every single day incase some roads move around overnight you still aren't really knocking my socks off with the idea that hiring 6024 people to support some transport infrastructure is some crazy high number in a city that supports 50,000 cab drivers. If you ever invent an advanced car that can drive 5 or 6 miles in a day you are pretty solidly in the realm that a whole united states scan can be done monthly with a workforce smaller than one amazon warehouse.
|
# ? Aug 29, 2018 06:12 |
|
Owlofcreamcheese posted:Okay? Even if you go to some absurd extreme where you need a separate employee for every single mile in new york city and you need to map the city every single day incase some roads move around overnight you still aren't really knocking my socks off with the idea that hiring 6024 people to support some transport infrastructure is some crazy high number in a city that supports 50,000 cab drivers. You're ignoring all the other people who have to handle and verify the data. You're ignoring that it's clear there will be hundreds of thousands of people involved all around to direct, operate, and maintain and interpret the equipment and data. You're ignoring how they fail to keep up today with much simpler street view scans.,which uses quite a bit of people already.
|
# ? Aug 29, 2018 06:27 |
|
fishmech posted:You're ignoring all the other people who have to handle and verify the data. You're ignoring that it's clear there will be hundreds of thousands of people involved all around to direct, operate, and maintain and interpret the equipment and data. The US has: 1.6 million truck drivers, 800,000 delivery operators,180,000 taxi drivers, 160,000 Uber drivers, 500,000 school bus drivers, and 160,000 transit bus drivers. Even in this weird fantasy hogwarts world where every single road totally reconfigures every single night and needs to be hand scanned daily so you need 260,000 people (one person to hand draw a loving sketch of each 10 mile section of road every single day because not even one lane of any road in the US stayed put for even two days) that is still not wildly out of sync with the number of people employed in other transportation infrastructure related jobs. You are really not establishing how even the most absurd and hyperbolic requirements for road scans makes any sort of number that is impossibly high.
|
# ? Aug 29, 2018 12:33 |
|
b-b-but the cost is nonzero
|
# ? Aug 29, 2018 12:57 |
|
BENGHAZI 2 posted:the Tesla would burn up with you inside it after you left the parking lot, and if it didn't you'd go to the wrong place Unless Tesla licenses their tech to makers of less expensive cars, and/or manages to start producing the $35k version of the Model 3 within the next couple years. It all comes down to which happens first: LIDAR units coming way way down in price, or Tesla perfecting camera-only self-driving.
|
# ? Aug 29, 2018 13:05 |
|
luxury handset posted:humans are generally ok drivers. many people go years if not decades without an accident. a smaller number of people cause a larger number of fatal accidents - mostly young people, old people, and people with bad driving habits
|
# ? Aug 29, 2018 13:42 |
|
Owlofcreamcheese posted:The US has: 1.6 million truck drivers, 800,000 delivery operators,180,000 taxi drivers, 160,000 Uber drivers, 500,000 school bus drivers, and 160,000 transit bus drivers. And? Most of those people can't be spared to go to random locations, because they already have jobs. Yes this is still wildly out of sync with the work that needs to be done to maintain the hyper accurate road maps required for Waymo's cars though. You are refusing to even comprehend how there's a difference between grabbing some medium res street view data that doesn't even stitch properly and the kind of high detail data needed for the cars.
|
# ? Aug 29, 2018 13:48 |
|
|
# ? Jun 11, 2024 00:41 |
|
Owlofcreamcheese posted:The US has: 1.6 million truck drivers, 800,000 delivery operators,180,000 taxi drivers, 160,000 Uber drivers, 500,000 school bus drivers, and 160,000 transit bus drivers. Yeah, that's only a little less than three times as many people as the entire Alphabet conglomerate employs right now. I'm sure that it would be no big deal for them to quadruple their workforce just for self-driving cars, and that it certainly wouldn't be a heavy cost that had a huge impact on self-driving car economics. If your plan for replacing human taxi drivers is to hire a comparable number of human map vehicle drivers plus skilled technical personnel to manually review and hand-tweak the maps, then you're not exactly going to end up saving on labor costs.
|
# ? Aug 29, 2018 14:17 |