Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Self-Driving Cars Must Meet 15 Benchmarks in U.S. Guidance

bloomberg.com

Post reply on HN
Print view View original

201–210 of 300 posts

Re: Self-Driving Cars Must Meet 15 Benchmarks in U.S. Guidance

#201

Post by blazespin »

chc wrote: post #195
Animats wrote: post #131

From the regulations: "Fall back strategies should take into account that—despite laws and regulations to the contrary—human drivers may be inattentive, under the influence of alcohol or other substances, drowsy, or physically impaired in some other manner." NHTSA, which, after all, studies crashes, is being very realistic. Here's the "we're looking at you, Tesla" moment: "Guidance for Lower Levels of Automated Vehic…

I don't understand this anti-autonomy cheerleading. It's like people on HN live in a parallel universe where there have been a bunch of deaths from cars running Autopilot, whereas in the world I live in, it seems to be somewhat safer than a human alone. Like, people can mess up either way, but they seem to be less likely to do so when the car is also looking out for them. What am I missing?

You have to compare the one death using autopilot to one death of people driving Teslas without autopilot. Musk tried to compare it against the universe of drivers (Teslas, kids driving crappy cars, etc), which was a complete false comparison.

So the reason it was a big deal is because it was a huge fatality. Tesla drivers are generally a pretty safe bunch. Statistically, if autopilot hadn't been engaged, that death would not have occurred. Autopilot makes Tesla drivers less safe, not more safe.

Also, the government is doing self driving industry a huge favor. These fatalities could screw over the whole industry if they get out of hand. Musk is giving self driving a bad name.

Re: Self-Driving Cars Must Meet 15 Benchmarks in U.S. Guidance

#202

Post by princeb »

chc wrote: post #195
Animats wrote: post #131

From the regulations: "Fall back strategies should take into account that—despite laws and regulations to the contrary—human drivers may be inattentive, under the influence of alcohol or other substances, drowsy, or physically impaired in some other manner." NHTSA, which, after all, studies crashes, is being very realistic. Here's the "we're looking at you, Tesla" moment: "Guidance for Lower Levels of Automated Vehic…

I don't understand this anti-autonomy cheerleading. It's like people on HN live in a parallel universe where there have been a bunch of deaths from cars running Autopilot, whereas in the world I live in, it seems to be somewhat safer than a human alone. Like, people can mess up either way, but they seem to be less likely to do so when the car is also looking out for them. What am I missing?

i don't see any anti-autonomy in that post?

he's definitely anti "disguising level 2 as autonomy" though.

Re: Self-Driving Cars Must Meet 15 Benchmarks in U.S. Guidance

#203

Post by MertsA »

blazespin wrote: post #199
Animats wrote: post #131

From the regulations: "Fall back strategies should take into account that—despite laws and regulations to the contrary—human drivers may be inattentive, under the influence of alcohol or other substances, drowsy, or physically impaired in some other manner." NHTSA, which, after all, studies crashes, is being very realistic. Here's the "we're looking at you, Tesla" moment: "Guidance for Lower Levels of Automated Vehic…

"Automatic driving will not be reached incrementally." Not true.. There are other ways of doing this incrementally. For example, slow speeds, closed roads (no pedestrians or other cars), only in good weather, ideal conditions, etc.

What he's saying is that it's not good enough to be able to control the car 90% of the time. Either it needs to be robust enough to operate safely without human intervention 100% of the time or it needs to somehow enforce that the driver is alert and capable of taking over 100% of the time.

We can't have an autonomous car that expects a driver to take over in a dangerous situation if that driver hasn't had to maintain control the entire time. For instance, there are youtube videos of drivers moving to the passenger seat in a Tesla with autopilot on.

Re: Self-Driving Cars Must Meet 15 Benchmarks in U.S. Guidance

#204

Post by drivingmenuts »

owyn wrote: post #188
ticklemyelmo wrote: post #109

Earlier quoted context omitted.

The cynic in me immediately interpreted "Post-Crash Behavior" as the exact opposite priority from "Privacy".

That's a very short section, but it looks more like how the driving system itself responds if a sensor is damaged in a crash. Basically, that it should hand control back to the driver. And also if you crash and then repair the system, it must somehow be validated/tested before being put back in service. The data collection "black box" side of it is in a different section.

> it must somehow be validated/tested before being put back in service

I pity the mechanics for that one. You just know the car manufacturers are not going to want some unwashed shade tree mechanic, or even a legitimate independent garage, to have access to do that.

Re: Self-Driving Cars Must Meet 15 Benchmarks in U.S. Guidance

#205

Post by pdeuchler »

e40 wrote: post #18
gumby wrote: post #6

Earlier quoted context omitted.

> I'm astounded that it seems like these regulations are going to be sensible... Was that hyperbole? I would say the majority of regulations (at least in OECD countries) are sensible, and many that are not are intended to be, are outdated, or are politicized.

Many people automatically assume a government can't make up sensible regulations. There are a lot of them in the US. It's a meme you hear all the time, especially in a POTUS election year.

You're right, we probably should be celebrating our lead tainted water, fracking, telecommunications monopolies, $300 Epipens, unaccountable financial institutions, pipelines across native land, corrupt campaign contributions, infrastructure decay, lack of decent health care...

Re: Self-Driving Cars Must Meet 15 Benchmarks in U.S. Guidance

#206

Post by mapt »

Animats wrote: post #131

From the regulations: "Fall back strategies should take into account that—despite laws and regulations to the contrary—human drivers may be inattentive, under the influence of alcohol or other substances, drowsy, or physically impaired in some other manner." NHTSA, which, after all, studies crashes, is being very realistic. Here's the "we're looking at you, Tesla" moment: "Guidance for Lower Levels of Automated Vehic…

I'm not sure they understand just how many intermediate steps requiring non-obvious technical progress are going to be required between 2, 3, 4, and 5. On a premapped track, 3 is just fine at present, but Google is nowhere near 3 for the sort of adverse conditions and unmapped road alterations that are common in much of the road network. This is going to be an iterative design process if it's moving forward at all.

A likely future is one where automation is only enabled for consumers as an option on a minority of roads (starting with the Interstate Highway System) that have been heavily mapped and managed, and we work from there, developing the algorithms at high sample size, then slowly extending out into the state highways and arterials. The roads and maintenance actions will likely also, as the tech progresses, have some modifications made to increase reliability.

These cars are going to need a large quantity of sensors; The Uber self-driving car has "something like 20 cameras, a 360 degree radar, and a bunch [7] of laser [rangefinders]", and this is a decent start; a Tesla and even a Google car is simply not equipped for enough edge cases to let a consumer near without making them hands-on-wheel liable to take over.

Re: Self-Driving Cars Must Meet 15 Benchmarks in U.S. Guidance

#208

Post by ryandrake »

drivingmenuts wrote: post #204
owyn wrote: post #188

Earlier quoted context omitted.

That's a very short section, but it looks more like how the driving system itself responds if a sensor is damaged in a crash. Basically, that it should hand control back to the driver. And also if you crash and then repair the system, it must somehow be validated/tested before being put back in service. The data collection "black box" side of it is in a different section.

> it must somehow be validated/tested before being put back in service I pity the mechanics for that one. You just know the car manufacturers are not going to want some unwashed shade tree mechanic, or even a legitimate independent garage, to have access to do that.

Let alone self-maintenance. There goes the right to repair...

Re: Self-Driving Cars Must Meet 15 Benchmarks in U.S. Guidance

#209

Post by Animats »

rhino369 wrote: post #178
marricks wrote: post #169

Earlier quoted context omitted.

For the Tesla comment, what more could you expect from them right now? If you have auto pilot on your hands have to be on the wheel or it vibrates and then it starts slowing down e eventually since it knows it's not safe. People can work around any system for this, but stuff like this makes it have to do it consciously. Seems pretty reasonable on Tesla's part.

They don't have to implement SAE 2 and if their implementation isn't safe, doing the best they can might not be good enough. I think there may be a fundamental flaw with lane keeping. It removes the driver from doing anything but still requires constant vigilance. That might be asking too much. My ADD is too strong to wach the road without having to do any part of the driving. I suspect a lot of people are the same w…

There's a good chance that's what happens. Auto-braking (level 1) is likely to become standard, like anti-skid braking. Full automatic driving (levels 3-5) will be options. Semi-automatic steering (level 2) may disappear as the higher levels start to work. The shared responsibility between driver and control system is too messy at level 2.

Re: Self-Driving Cars Must Meet 15 Benchmarks in U.S. Guidance

#210

Post by CamperBob2 »

blazespin wrote: post #199
Animats wrote: post #131

From the regulations: "Fall back strategies should take into account that—despite laws and regulations to the contrary—human drivers may be inattentive, under the influence of alcohol or other substances, drowsy, or physically impaired in some other manner." NHTSA, which, after all, studies crashes, is being very realistic. Here's the "we're looking at you, Tesla" moment: "Guidance for Lower Levels of Automated Vehic…

"Automatic driving will not be reached incrementally." Not true.. There are other ways of doing this incrementally. For example, slow speeds, closed roads (no pedestrians or other cars), only in good weather, ideal conditions, etc.

There are other ways of doing this incrementally.

Yes, and they're all undesirable, unworkable, or useless, as your own post points out.

Post reply on HN