Hacker News Reader

Comment by Someone1234 | Hacker News Reader

If your point is "Autopilot only works on perfect roads" then they might have well pull the plug right now.

Dirt, snow, rain, fog, dark, and other common conditions could obscure even correctly painted lines, and it just isn't realistic to expect perfect roads throughout such a large network anyway (in particular as human drivers can infer the line is meant to be there from the neighboring lines).

At its core, a system like Autopilot is meant to stand in for humans. Plus in this video where they reproduced the accident there was a full chevron pattern so the argument is flawed/wrong anyway: https://www.youtube.com/watch?v=6QCF8tVqM3I

View on HN

Man, that's incredibly bad. I drive that route all the time and it's not even remotely difficult to follow the highway there.

Considering Tesla's must have driven through there hundreds of thousands of times, they should have data to make sure the car at least follows the highway. I don't crash in to a barrier when a paint stripe gets scuffed in between my Monday morning commute and my Tuesday morning commute.

I'm curious if they have data about how often users need to take over from autopilot in that area. It seems like this shitty behavior was easily reproducible, so...

It's not so simple - you can't simply follow the highway as you have done hundreds of thousands of times, because if the paint stripes have changed since Monday, then these might also be new (or temporary) stripes painted by a road construction crew overnight and you now must follow the new stripes.

Now imagine all people have Tesla, and someone makes a prank on the middle of a bridge, they paint a line that makes all traffic plunge down, imagine now that they are so confident they removed the driver seat, you have all those passengers looking on the windows on how their cars are driving them to their deaths and there is nothing they can do.

So should this cars follow some line blindly?

I think that this tech should stil be called a drive assist, and assist the driver, Tesla rushed it in enabling the autopilot thing, I am sure if you compare the stats with similar cars(expensive,safe and with drive assist) the stats will be against autopilot, I do not understand why they had to rush the autopilot and not keep it as a driver assist, to warn you and prevent some crashes)though it seems it can't protect from frontal collisions with static objects)

You could say the same for trains though - they follow a track in a similar sense to Tesla's autopilot. Someone could perform a "prank" by putting something on the track in order to derail a train. How often do you see that happen? Painting a line on the road is obviously easier to do, and could also happen by accident, but I think the chances of any danger are extremely low.

Hopefully when we start removing the driver from cars with autopilot, they will be following some sort of under-road NFC-like guidance system rather than painted lines.

There’s a theory out there that a penny on a rail can derail a train. You wouldn’t think people would do such a thing, with such disastrous consequences... but they do. Thousands of times a year.

Actually derailing a train takes serious effort; but people still try constantly. Derailing a Tesla seems somewhat easier, so I guess we’ll see.

On youtube there are old black and white films produced to show the result of rail sabatouge. Even with a foot cut from the track the train did not derail. Quite impressive. I used to flatten coins as a kid all the time by leaving them on the tracks.

It is much harder/expensive to place a big truck on heavy object on a train track then painting a line. The trains also have a driver that will try to stop the train so reduce the impact.

This self driving cars can be tricked by bad conditions or some optical illusions, like you have a t-shirt with some pattern on it and the Tesla car will coma and hit you

Agreed. That was in Chicago. It also was reproduced exactly in CA crash location in https://electrek.co/2018/04/03/tesla-autopilot-crash-barrier...

Indeed. His Tesla was clearly following the clearest left-side white line. Which unfortunately pointed the car right at the gore point.

But what amazes me is that Tesla autopilot apparently won't swerve or stop when it sees an unmoving object straight ahead. Whether it's a semi, a fire truck, or a gore point. But I gather that trying to detect that stuff generates too many false positives. And that overall, it would arguably cause more accidents net.

Edit: I have faced that ambiguity myself. When driving, perhaps foolishly, in fog so dense that I could only see the lines on the road.

But what amazes me is that Tesla autopilot apparently won't swerve or stop when it sees an unmoving object straight ahead.

It won't even brake unless the thing in front looks like the back end of a car or truck. Even then, if it's stationary, it may still get hit. It won't swerve, ever; it has no logic for that. Four times that we know of, a Tesla has plowed into a vehicle partly blocking a lane.

Putting something on the road which pretends to self-drive and can't detect solid obstacles is criminal.

Will it brake for pedestrians?

If I was walking, I sure would not trust it to. But their feature is for the highway, where peds are not expected. I am sure they expect to shift blame by that reasoning.

I doubt that it would.

But seriously, name "autopilot" aside, it's intended for divided highways. With high-quality road markings. And no stop lights.

And to be fair, airplane autopilots handle flying and landing. They don't handle driving between runways and gates.

Both Tesla and Waymo are at too low an abstraction level for resilience. Cars need to understand roads/walls as concepts and make common sense inferences.

They're not comparable at all. Google/Waymo uses lidar and has been at it since 2009 and per https://waymo.com/faq they are "fully automated vehicle" but alas not for sale yet.

Tesla's self-driving vaporware being sold since Oct'16 is camera only and should be activated any day now ?

https://twitter.com/elonmusk/status/823632597284691969 " Yes, safety should improve significantly due to autonomy features, even if regs disallow no driver present 23 Jan 2017"

3 months maybe, 6 months definitely 23 Jan 2017

Waymo’s approach is also brittle, e.g. to changes in the road since the last map pass.

Tesla also uses radar and ultrasonic. Lidar isn’t necessary, just makes depth measurement easier. It’s lower res than cameras, low range, power-inefficient, conflicts with other lidar. Humans drive fine in sunlight, without a coherent IR scanning beam.

better safe than sorry, and I doubt the car seizes up and bails

How do you know how much Waymo's approach rely on maps vs. vision? My expectation would be that maps are really only used for "macro" navigation but anything beyond that (lanes, avoiding obstacles, etc.) would be done on the spot.

Do you have any public source detailing this or do you just make this claim based on your own assumptions?

(Disclaimer: I work at Google but have no idea how our self-driving cars work.)

https://medium.com/waymo/building-maps-for-a-self-driving-ca...

> Of course our streets are ever-changing, so our cars need to be able to recognize new conditions and make adjustments in real-time. For example, we can detect signs of construction (orange cones, workmen in vests, etc.) and understand that we may have to merge to bypass a closed lane, or that other road users may behave differently.

Doesn't sound like the cars blindly trust potentially stale maps.

That makes more logical sense.

The low sunlight is hard to deal with by humans and computer vision isn't better. In the meantime, Tesla's CEO doesn't believe Lidar is necessary (https://www.theverge.com/2018/2/7/16988628/elon-musk-lidar-s...).

My advice to the company BOD & shareholders: have him take the Udacity's self driving car online course and after the first project, he'll clearly understand how limited detecting lanes by CV only can be. To Uber's ex-CEO's point: LIDAR is the SAUCE.

Musk's central point is that if humans can drive with two cameras, so can a machine. And he's right. Why wouldn't we do just as well as the visual cortex?

Humans have superior intelect, even if specific performance is lower. We have experince. We learned to drive, in the particular area where you operate your car generally, with all its idiosyncorcies. Then consider eye contact, nonverbal communication, bias, and personal investment in outcome that computers are incapable of. Its not as simple as better sensors and reaction time.

But isn't the whole point of autonomous cars that humans are pretty shitty drivers? If I could augment my vision with a 360º setup of cameras and LIDAR you better believe I would!

I'd say people are actually pretty good drivers. I'm more interested in autonomous for the time savings than I am for the potential safety improvements.

Tens of thousands of people are killed every year[1] in just the United States, humans are awful at driving. If autonomous vehicles are able to make driving as safe as flying it will be like curing breast cancer in terms of lives saved.

[1] http://www.nsc.org/NewsDocuments/2017/Fatality-estimates-Jun...

There are hundreds of millions of people in the US. Most of them drive. Is a few tens of thousands fatalities that bad in an absolute sense?

Are you serious? Yes, a stadium full of people dying every year is "that bad" in an absolute sense.

By what metric? You're leaving out any sense of proportion.

our eyes are a lot better than cameras in a lot of ways. Eyes have better dynamic range, better sensitivity in low light, extremely high resolution in the center, and an extremely wide field of view. The nerve cells in our retina are also wired up to do a lot of processing in real time at extremely high resolution, eg. things like motion/zoom/edge detection.

And that‘s not even taking into account that we actually understand what we see and can reason about unexpected input and react accordingly.

>> They're not comparable at all. Google/Waymo uses lidar and has been at it since 2009 and per https://waymo.com/faq they are "fully automated vehicle" but alas not for sale yet.

If it's already fully automated, then why is it not for sale, yet?

Perfect it before selling vaporware or risking it like Uber and Tesla. They are partnering with Manufacturers already e.g.

https://www.theverge.com/2018/4/2/17189764/waymo-honda-self-...

https://www.theverge.com/2018/3/27/17165992/waymo-jaguar-i-p...

The required hardware is too expensive to make it commercially feasible.

Is it ? It's down to 8k https://arstechnica.com/cars/2017/01/googles-waymo-invests-i...

> If your point is "Autopilot only works on perfect roads" then they might have well pull the plug right now.

It's even worse: wait until someone does something malicious to an autopilot car to cause it to crash/kill.

I hate to be so cynical, but if you're developing this stuff, you have to imagine worst case scenarios. And there are much much worse scenarios when malice is involved.

https://xkcd.com/1958/

Most people aren't murderers. On the other hand, con artists when the expected risk to marks is minimal...now we're talking!

I've been had by local traffic cops in $COUNTRY, just a few km after crossing the border: two lanes in one direction, separated by a solid line ("no changing lanes"), running parallel for about a km. Left lane suddenly ends - asphalt gradually narrows from the left until there's only the right lane left. Your options are a) crossing the solid line into the right lane, and...wait, that's all, there's no way to go off the road. And guess what, traffic cops conveniently positioned right there: "you have crossed the solid line, that's a ticket and XYZ in local currency" "but there's nothing else I could have done!" "we don't care, you crossed the solid line!" Repeat until ticket is paid - note that this is obviously targetted at people who are not likely to fight the ticket (and presumably a mistaken/illegal road marking) in local traffic court.

I could imagine a very similar scenario set up to trip up SDVs (and fleece their presumably affluent occupants). The problem comes when con men underestimate the risks, and accidentally kill people.

they know this, and they're targeting harsh conditions, not perfect

I have been a consistent critic of Autopilot; so much so that my last post with that very same video got nuked hard for awhile.

The same suggestion as before, autopilot needs to be turned back into learning only. It never should have left. There were way too many videos just on youtube showing how bad it could be. It should simply be recording what it would decide to do versus what the driver did and flag every exception. It certainly can see barriers and its routines could save every instance where it might have hit something.

The NHTSA really should demand that Autopilot features be disabled. There are even cases of it failing to park without hitting objects.

Musk should be lauded for his EV push, it opens doors to all sorts of new technologies being incorporated in cars and will one remove the need for gasoline/diesel as the primary drive in all cars. However he needs to learn to step back on technologies that are clear cutting edge and not ready for prime time.

At this time features like "AP" should be used not to take over for the driver but instead assist the driver by keeping them from being unsafe. This means doing similar to what blind spot monitors to and lane drift systems work on.

Reply on news.ycombinator.com