|
We should identify drivers who are worse than today’s autonomous cars and make it illegal for them to drive themselves. Then we should identify the drivers who are better and make it illegal for them to cede the wheel.
|
# ? Oct 27, 2023 01:03 |
|
|
# ? Jun 1, 2024 17:30 |
|
Vegetable posted:Waymo’s driverless cars experienced 76% fewer accidents than human-driven cars based on a recent study using insurance data. That study was run and released by Waymo themselves, who have not yet submitted the data behind that study to independent peer review. Given that Waymo's cars haven't been operating completely driverless for very long, and operate only in two cities, it's quite early to say that this is good data. As for the rest, this is just totally unsupported opinion. You say that human drivers are awful (compared to what???), then say that robot drivers are "fairly conservative" (are they??? compared to what???), and then say that there should be a presumption of superior safety (why?????). You're just throwing poo poo out there with no backing whatsoever and acting like it's fact. But let's move away from the theoretical ideal of potential robot cars, and look at the actual driverless cars and how they're behaving on our streets right now*: https://www.sfchronicle.com/projects/2023/self-driving-cars/ quote:San Francisco firefighters were battling a two-alarm apartment blaze on the corner of Hayes and Divisadero streets during a recent Sunday morning when a driverless Cruise car entered the active firefighting scene and nearly ran over fire hoses on the street. * Technically, Cruise doesn't have driverless cars on the streets anymore, because they got their driverless license suspended. Officially, it was suspended for poorly handling a situation with an injured pedestrian on the road, but the real reason for the suspension is almost certainly because the company intentionally concealed information from regulators, cutting off the video they provided before the part where the stopped car started driving again with the pedestrian still trapped underneath and dragged them 20 feet down the road before stopping again.
|
# ? Oct 27, 2023 01:12 |
|
Platystemon posted:We should identify drivers who are worse than today’s autonomous cars and make it illegal for them to drive themselves. Then we should identify the drivers who are better and make it illegal for them to cede the wheel. What if we could calibrate the ai to exactly match each passengers' safety profile? Or charge them based on the gap in safety between them and self-driving default?
|
# ? Oct 27, 2023 01:26 |
|
Main Paineframe posted:
|
# ? Oct 27, 2023 02:08 |
|
Kwyndig posted:I'd think that even the implication that driverless cars are dragging pedestrians down the street after running them over would be enough to scuttle the program until they can drat well prove it won't happen again. The running-over part was probably unavoidable - another car hit the pedestrian and knocked them directly into the path of the driverless car, which slammed the brakes but was too close to stop in time. The part that was the driverless car's fault is that after stopping, the car started moving again, driving down the road with the person still trapped underneath as it looked for a good place to pull over. I feel like if that was all there was to it, they probably would have gotten away with a slap on the wrist as long as they apologized and promised it wouldn't happen again. The problem is that Cruise only told the DMV and the press about the first part, the part that wasn't their fault. The videos they provided cut off right when the car first came to a stop. California regulators only found out about the rest of it a couple weeks later when they heard the full story from a NHTSA investigator, and that got them pissed off.
|
# ? Oct 27, 2023 02:55 |
|
And this kind of lovely behaviour perfectly encapsulates why these companies should not be trusted to sling around robotic death machines on our roads, regardless of what forums poster Electric Wrigglies claims. Thankfully European car manufacturers are taking things more slowly and doing things like, oh I don't know, rigorous testing on closed roads *where they don't kill people*. Companies like Cruise and Tesla are shamelessly lying about the capabilities of their cars and the incidents they get into and we're supposed to be happy about it because "humans cause accidents, too"? That doesn't excuse these companies' insanely lovely and reckless behaviour. Sagacity fucked around with this message at 06:35 on Oct 27, 2023 |
# ? Oct 27, 2023 06:32 |
|
Main Paineframe posted:No, what they are saying is "We need to have stringent safety standards and have companies pass those standards before being allowed to do a small pilot program, and then eventually grow that program over time in a way that keeps the public safe". That's a very different thing from what you're rewording their post to say, and I'm not sure why you tried to rephrase the post, because the original wording is very straightforward and easy to understand, and I don't think it's a particularly unreasonable stance either. Your post makes it seem like Waymo is just putting its cars on the market for general release instead of being a part of many small and limited test periods and pilot programs that it is. Not much different to clinical trials of medicine prior release into the general use (clinical trials, especially mental medication - also put non-trial people at risk). Like a clinical trial where a fellow goes weird and kills his workmate, the Waymo program was paused/stopped to look into why it happened etc. That the company was weird about it as you point out is definitely worthy of further scrutiny and sanction by the regulators. You posted a list of anecdotal incidents that almost certainly have multiple equivalents every day with driver driven cars across the States. It is demonstrating that there is still work to be done to improve driverless cars even further and hence why they are not in general release. I expect they will get better. Bit like how Deep blue struggled to beat Kasparov in 1997 but now it is trivial for computers to beat the words best Chess and even Go players. The list for me mostly looks like punch list work, expected hiccups with commissioning anything new of the scale of what is being attempted. And finally, the US is not the only place thinking about this. Driverless cars are legal in Shenzhen, China and roll outs through China will continue to advance the state of art as the automation is fine tuned to what is likely a more challenging driving environment. Maybe the US can ban development until it is demonstrated conclusively enough in real China use for all but the covid deniers. What will happen then is the rest of the world will buy Chinese tech and not US tech due to the lack of development of US versions.
|
# ? Oct 27, 2023 10:39 |
|
Electric Wrigglies posted:Maybe the US can ban development until it is demonstrated conclusively enough in real China use for all but the covid deniers. What will happen then is the rest of the world will buy Chinese tech and not US tech due to the lack of development of US versions. Ensuring that development of technology is done safely and with minimal risk to human life is not the same thing as banning development. Stop with the bad-faith false equivalencies.
|
# ? Oct 27, 2023 10:51 |
|
Maybe I'm wrong but when pharmaceuticals want to test experimental drugs on me they have to ask express permission. Which is not the case for driverless cars, so I don't think the comparison is very valuable.Electric Wrigglies posted:What will happen then is the rest of the world will buy Chinese tech and not US tech due to the lack of development of US versions. This is the level of argument I expect on fox news. Oh no heavens, not China!!
|
# ? Oct 27, 2023 10:53 |
|
Dirk the Average posted:Ensuring that development of technology is done safely and with minimal risk to human life is not the same thing as banning development. Stop with the bad-faith false equivalencies. Driverless cars for the most part have gotten as good as they are going to get without real world testing. Banning real world testing while China rolls out Autonomous Driving network (alongside its monster public rail network) absolutely without a shadow of a doubt is equivalent to banning development from keeping up with Chinese development. E) Mega Comrade posted:Maybe I'm wrong but when pharmaceuticals want to test experimental drugs on me they have to ask express permission. Which is not the case for driverless cars, so I don't think the comparison is very valuable. It may surprise you to know that people on clinical trials are not always banned from going out in public or from operating vehicles. People sometimes have really bad reactions to medications that result in others losing their life. Electric Wrigglies fucked around with this message at 10:56 on Oct 27, 2023 |
# ? Oct 27, 2023 10:54 |
|
Electric Wrigglies posted:Driverless cars for the most part have gotten as good as they are going to get without real world testing. Banning real world testing while China rolls out Autonomous Driving network (alongside its monster public rail network) absolutely without a shadow of a doubt is equivalent to banning development from keeping up with Chinese development. Bullshit.
|
# ? Oct 27, 2023 11:08 |
|
Electric Wrigglies posted:
What the gently caress are you on about? Experimental medical testing is not in any single way comparable to experimental testing of sending multi-ton steel objects driving through regular traffic.
|
# ? Oct 27, 2023 11:33 |
|
And when has a medical reaction killed somebody else. I want sources for this ludicruous claim.
|
# ? Oct 27, 2023 11:39 |
|
Electric Wrigglies posted:(clinical trials, especially mental medication - also put non-trial people at risk). You’ve been watching too much Serenity or Gen V or too many D.A.R.E. presentations. That’s not how mental illness works, and it’s not how treatment for mental illness works.
|
# ? Oct 27, 2023 12:00 |
|
Electric Wrigglies posted:Your post makes it seem like Waymo is just putting its cars on the market for general release instead of being a part of many small and limited test periods and pilot programs that it is. Not much different to clinical trials of medicine prior release into the general use (clinical trials, especially mental medication - also put non-trial people at risk). Mental medication?! e: Sorry - that was flippant. But I would like to see examples.
|
# ? Oct 27, 2023 12:40 |
|
Preventable medical mistakes was killing, in the late 90's, in excess of 44k of Americans each year, leading vehicle accidents (~40k) as the number one cause of preventable deaths and you lot are completely convinced that never in the US has someone that is on a clinical trial ever (for example) made a mistake driving a vehicle or some other action (handling firearms, workplace equipment, completing medical procedures, etc) leading to the death of another person. To quote my esteemed poster; Dirk the Average posted:Bullshit. Adverse unintended outcomes of clinical trials is routine. So much so that there is a body of study work dedicated to if and how to appropriately compensate such outcomes. Sometimes these outcomes are serious and sometimes even leading to the death of participants, their unborn children and (quite infrequently), others. That some people might die in the quest for better medication and procedures is the price of doing business that the public tolerates (even as we all expect improvement over time). That there is a lack the imagination to consider that amongst the 10's of thousands of preventable deaths worldwide that there would be weird edge cases (such as someone on a clinical trial falling asleep at the wheel of a car due to medication caused drowsiness) is probably why there is demand for unprovable perfection prior to an AD trial (how to prove AD cars will be satisfactorily safe without going onto a public road? You can't, so therefore it can never be considered safe enough to trial). Incidentally, this demand for perfection is also why US pilots are automatically (at least temporarily) banned from flying if they declare or admit to a mental condition or depression (leading to pilots self-diagnosing and self-treating depression amongst other outcomes). Even if it is perfectly treatable and the pilot is managing that illness, their career is hosed through the abundance of caution that prefers pilots self-treating for depression rather than flying while observing professional treatment.
|
# ? Oct 27, 2023 13:27 |
|
Edit: Actually you know what? Let's drop the medical comparison. It is quite clear that we aren't going to agree on that comparison and we'll just end up shouting past each other forever. Let's go with the comparison to airline automatic pilot implementation. Those implementations have certainly made life a lot easier for pilots, but they've also led to a noticeable skill decay due to lack of constant muscle memory training. It has also led to a reaction time issue as the pilots have to deal with both the sudden issue and get themselves into the groove of controlling the airplane fully. I'm sure the driverless car will eventually be able to keep the engine running and the wheels turning in the right direction during normal operations, but that will also lead to a general skill decay in the people ostensibly in control of the vehicle in case an emergency props up or the driverless function fails for whatever reason. Having to react to an incident near the road will potentially require the driver to know how to handle their car off road or during awkward angles of towing a vehicle, pushing objects away from pedestrians, placing the car between an enraged animal and a person, getting out of a flooding situation before the water rises too high, avoiding falling rocks that are still too high enough up for the car to register. In the end, not all cars will be useful as driverless vehicles, since the operator will be using the car as part of their job. SerthVarnee fucked around with this message at 13:51 on Oct 27, 2023 |
# ? Oct 27, 2023 13:36 |
|
Electric Wrigglies posted:To quote my esteemed poster; Again, the "Bullshit." I stated referred to: Electric Wrigglies posted:Driverless cars for the most part have gotten as good as they are going to get without real world testing. Banning real world testing while China rolls out Autonomous Driving network (alongside its monster public rail network) absolutely without a shadow of a doubt is equivalent to banning development from keeping up with Chinese development. Which is, frankly, bullshit. It's even worse when you continue to put words in my mouth by saying that I am advocating for "banning real world testing." Regulation is not a ban. As much as I don't want to lead into a "China bad" thing, my anecdotal experience in the medical industry has been that there is a certain lack of care about safety in a lot of their testing (multiple experiences across multiple companies). Bribing the official that is there specifically to stop bribes is a thing, and you can fast track therapies that really aren't at the point where they should be deployed on patients. Citing China as a case study in how we should be approaching research that has a cost in human lives is not ideal, to say the least, according to my limited experience. You have done a poor job of convincing me (and others) that we need to just allow companies to throw AD vehicles on the road. Pilot studies and going through proper regulatory channels are things that should absolutely be required for companies that plan to do this. I am also unconvinced that real world testing is required for progress at this point - while that may be true for the most cutting edge programs, it is certainly not true across the entire industry. And those programs that are at the point where they can be unleashed upon the public should be doing so with extreme caution that follows a strict protocol. We are certainly not at the point where these vehicles should be available for commercial use, either (as evidenced by Cruise is recent months).
|
# ? Oct 27, 2023 14:04 |
|
Electric Wrigglies posted:Your post makes it seem like Waymo is just putting its cars on the market for general release instead of being a part of many small and limited test periods and pilot programs that it is. Not much different to clinical trials of medicine prior release into the general use (clinical trials, especially mental medication - also put non-trial people at risk). First off, it wasn't Waymo, it was Cruise. Second, the program wasn't stopped when a needless injury happened - it was only stopped when regulators found out that the company had actively attempted to deceive them, which I don't believe is a normal feature of clinical trials. I don't know why you keep talking about "general release". I thought my stance, and the stance of the article, was fairly clear: that these cars are not ready to be on public roads. Yes, I'm sure they'll get better over time. But they should be withdrawn from public roads until they're at least better enough to stop slamming into emergency vehicles all the time. And they certainly shouldn't be taking passengers. I don't believe we should be striving to emulate the Chinese government's safety practices. It's not a government that's exactly known for its strict safety regulation of unscrupulous companies. Nobody is talking about banning development, they're talking about stopping public alpha tests on public roads. Electric Wrigglies posted:Driverless cars for the most part have gotten as good as they are going to get without real world testing. Banning real world testing while China rolls out Autonomous Driving network (alongside its monster public rail network) absolutely without a shadow of a doubt is equivalent to banning development from keeping up with Chinese development. There is absolutely no way that's true. It is definitely possible to improve them further without having them on public roads. Having them on public roads may speed up development by making all the developers' mistakes and bad assumptions very obvious, but that's hardly worth the risk to public safety. In fact, I'd go so far as to say that these companies are putting these cars on the street before they're really ready to do so, not for technical reasons but rather for business reasons. With a bunch of companies all competing to be the first ones to break into a totally new market for totally new technology, there's a strong incentive to rush things.
|
# ? Oct 27, 2023 14:21 |
|
Oxyclean posted:
That's the joke. The only way is going to happen is over forty years, as insurance companies lobby to only have insurance count when the AI does the driving.
|
# ? Oct 27, 2023 16:27 |
|
Main Paineframe posted:In fact, I'd go so far as to say that these companies are putting these cars on the street before they're really ready to do so, not for technical reasons but rather for business reasons. With a bunch of companies all competing to be the first ones to break into a totally new market for totally new technology, there's a strong incentive to rush things. This is exactly why the vehicles are going on the road ASAP and tech companies keep Mister Facetious posted:That's the joke. The only way is going to happen is over forty years, as insurance companies lobby to only have insurance count when the AI does the driving. While also requiring the AI to immediately give control to the human in the car right before impact so they can deny claims.
|
# ? Oct 27, 2023 16:32 |
|
Platystemon posted:We should identify drivers who are worse than today’s autonomous cars and make it illegal for them to drive themselves. Then we should identify the drivers who are better and make it illegal for them to cede the wheel. This would unironically actually be a great place to start. We already have actuarial tables for car insurance that predict, accurately, whos likely to be a safe driver and who isn't. As things currently stand though being an unsafe driver just means a higher car insurance price premium - which means its just a measure against poor people. Rich assholes who drive horribly and get in lots of accidents can just pay their way out of it. If in a few years we could restrict everyone in say the highest 10%-20% of risk profile to AI cars it'd make a huge difference to road deaths.
|
# ? Oct 27, 2023 16:45 |
|
Mandatory breathalyzers in all cars too, since AI is probably already safer than drunks
|
# ? Oct 27, 2023 16:46 |
|
Mister Facetious posted:That's the joke. The only way is going to happen is over forty years, as insurance companies lobby to only have insurance count when the AI does the driving. Which is funny because it sounds like AI cars already have a habit of going "
|
# ? Oct 27, 2023 16:48 |
|
Evil Fluffy posted:This is exactly why the vehicles are going on the road ASAP and tech companies keep
|
# ? Oct 27, 2023 17:03 |
The real issues are not with the drivers in the US. Contrary to popular belief, Americans aren't any dumber than anyone else, and focusing on the driver as the root cause of an accident ignores the massive systemic failures that actually cause our high level of road fatalaties. Off the top of my head:
Safe autonomous cars can help with these issues, but not as much as actual public transit. And, autonomous cars don't help with the other harms from cars, such as
Autonomous cars are maybe a degree of harm reduction for cars, but ultimately, we need to move primarily to other forms of transit for most people, and we need to do so as soon as we can.
|
|
# ? Oct 27, 2023 17:22 |
|
If cars go full AI then it'd be incredibly cheap for people and businesses to deploy them and use them for all sorts of utility purposes. That would certainly lead to safer roads if the AI is just driving in bumper to bumper traffic all day.
|
# ? Oct 27, 2023 17:33 |
|
VikingofRock posted:The real issues are not with the drivers in the US. Contrary to popular belief, Americans aren't any dumber than anyone else, and focusing on the driver as the root cause of an accident ignores the massive systemic failures that actually cause our high level of road fatalaties.
|
# ? Oct 27, 2023 17:35 |
|
VikingofRock posted:
Assuming that eventually cars achieve legal self driving on city streets, which is a huge if, we're going to have to face that it's cheaper for an EV to just drive around the block at low speed for hours than pay for parking.
|
# ? Oct 27, 2023 17:41 |
|
Vegetable posted:American drivers are dumber, in the sense that Driver Ed is trash. In most developed countries the training is conducted by licensed professionals not your mom and dad, and the test requires more than just taking the car around the block. There are complex reasons why the bar is set so low, but it’s undoubtedly a big reason why its drivers are so awful. This does very from state to state at least. I recall a fellow from Germany who spoke about how the licensing requirements in various states related to what he had, and how some states had basically worthless licenses while others were good enough to transfer over easily.
|
# ? Oct 27, 2023 17:53 |
|
VikingofRock posted:
And then there are drunk drivers who still get restricted driving licenses because otherwise they can't hold a job and support themselves.
|
# ? Oct 27, 2023 18:41 |
|
VikingofRock posted:The real issues are not with the drivers in the US. Contrary to popular belief, Americans aren't any dumber than anyone else, and focusing on the driver as the root cause of an accident ignores the massive systemic failures that actually cause our high level of road fatalaties. The US in 2022 had 8.3 deaths per 1 million passenger km driven. In Australia, a country thats known for having both large cars and similar road design the figure is 5.2. In Canada, ditto, its 5.1. Thats a huge difference. Public transport doesn't impact this figure. In European countries its even better - 3.0 in Norway, 3.8 in Ireland, 3.8 in the UK etc. Driving tests in the US are notoriously easy, and driver quality low as a result - theres a reason very few countries will exchange US licenses for a local one. This is a typical first world country's list of states recognised as having valid driver education/skill levels to justify a license swap: quote:An EU or EEA country
|
# ? Oct 27, 2023 20:22 |
|
Electric Wrigglies posted:Driverless cars for the most part have gotten as good as they are going to get without real world testing. Banning real world testing while China rolls out Autonomous Driving network (alongside its monster public rail network) absolutely without a shadow of a doubt is equivalent to banning development from keeping up with Chinese development. Name a single reason why this real world testing must be done with: 1. human non-employee passengers 2. no human employee present, so that they can take over just on the off chance that, oh, I dunno, the AI drives over someone and then continues moving forward.
|
# ? Oct 27, 2023 20:35 |
|
Speaking of real world testing, Cruise just 'paused' taxi services in all markets. Waiting to hear that this is no true AI.
|
# ? Oct 27, 2023 21:04 |
|
Steve French posted:Name a single reason why this real world testing must be done with: The only reason is if you’re running out of money and need to use the cars to generate revenue. This piece does a great job explaining why Cruise made such a reckless decision: https://apperceptive.substack.com/p/i-knew-this-was-coming
|
# ? Oct 28, 2023 00:58 |
|
Rhetorical Funhouse, this Tech Nightmares thread has become.
|
# ? Oct 28, 2023 16:06 |
|
Vegetable posted:The study compared human-driven cars in SF/Phoenix to Waymo cars. It’s like for like. SF is probably one of the hardest cities to drive in in America. That may not be enough for you, but it’s enough for me and frankly it just confirms what safety experts have been assuming for the longest time. You can wait for a nationwide launch before you make your mind up though oh come on SF doesn't even have snow and has significantly less rain than many other US cities.
|
# ? Oct 29, 2023 22:52 |
|
Lyesh posted:oh come on SF doesn't even have snow and has significantly less rain than many other US cities. I wanna see Waymo handle Duluth in the winter.
|
# ? Oct 30, 2023 15:35 |
|
Deuce posted:I wanna see Waymo handle Duluth in the winter. Thinking back to the time in Pittsburgh when cold damp winds put a thin layer of ice on Carson street in about an hour or so and there is no loving way an AI is going to be able to handle that without ending up slamming into a building or careening out onto the train tracks (or overshooting those and ending up in the trees if not the river itself). I guess it's not like the AI will have any worse visibility in whiteout conditions on 79 or 90 than a person does. Maybe. Evil Fluffy fucked around with this message at 18:56 on Oct 30, 2023 |
# ? Oct 30, 2023 18:53 |
|
|
# ? Jun 1, 2024 17:30 |
|
Evil Fluffy posted:Thinking back to the time in Pittsburgh when cold damp winds put a thin layer of ice on Carson street in about an hour or so and there is no loving way an AI is going to be able to handle that without ending up slamming into a building or careening out onto the train tracks (or overshooting those and ending up in the trees if not the river itself). An AI trained extensively on those conditions would probably do better than the average human driver in those conditions.
|
# ? Oct 30, 2023 18:55 |