First pedestrian death by autonomous car...

[nctnico]

...
...
There is no reason why an AV should be any smarter than a hammer. After all an AV is a tool. A sophisticated tool but a tool nevertheless. An AV should stick to the traffic rules and (if possible) don't hit anything. Still want common sense? Put a monitoring device inside an AV so a (human) call center operator can deal with 'situations'.

We are actually on the same page.  My argument is not whether AV should have that capability or not.  Instead, I was arguing about our expectations.

You know to not to expect (in your words) AV smarter than a hammer, where as, many others expect AV to be rather smart, perhaps even smarter than human.

My point is, we should expect it to be a rather dumb tool lacking common sense.  That deficit in turn makes it to be lacking the real intelligence of a real driver.  It will therefore will make mistakes that human with our common sense could easily avoid.

We are not on the same page. Common sense is different between people. Also many people don't react at all in a panic situation. I used to work at a research institute for human behaviour and it is mind boggling how many people lock up or do the wrong thing in life threatening situations. An AV OTOH will always do as it is programmed. It cannot get distracted or do stupid things like humans do and that is a big win. You are putting way too much confidence in the ability of human drivers. It doesn't take a sophisticated AV to drive better than the average human driver.

[Rick Law]

...
We are not on the same page. Common sense is different between people.
...

I thought we both felt AV is not as smart as some expect...  Our reasons may differ, but we are both thinking that.  Are we not?

[raptor1956]

Apparently, when Uber switched from the Ford Fusion test cars to Volvo they also scaled back the number of lidar sensor from 7 to 1.  The lidar units scan 360 degrees, but the vertical angle is small so there are blind spots.

https://www.reuters.com/article/us-uber-selfdriving-sensors-insight/ubers-use-of-fewer-safety-sensors-prompts-questions-after-arizona-crash-idUSKBN1H337Q


I said it before and I'll say it again ... Uber is a bad company, rotten to the core!


Brian

[helius]

It doesn't take a sophisticated AV to drive better than the average human driver.

I've heard this repeated like clockwork every time the safety issue comes up. But an AV system that cost over a billion dollars has killed a pedestrian after roughly 2 million vehicle-miles. The average human driver would have to drive half a billion miles to kill a pedestrian (12.5 fatalities per billion vehicle-miles in the US in 2016, of which 15% were pedestrians). All self-driving systems put together have only done 20 million miles or so, with multiple fatalities. Are you calling systems with many billions of dollars of development costs unsophisticated?

I think at this point it's safe to retire "AVs don't have to be very good to be better than humans" as a discredited argument.

[Halcyon]

Also, on another point, since this was essentially a "test vehicle", I'm curious whether or not Volvo's own safety systems (city safety etc...) were disabled and the control of the car was relying solely on third-party sensors, cameras, software etc... Volvo's technology (whilst still being a computer and not 100% fool proof) is designed to detect and avoid exactly this kind of collision. Those cars are also fitted with pedestrian air bags which lift the bonnet and deploy around the windscreen. There did not appear to be any evidence of this system activating.

Now that is an interesting question.

Called it!

I don't know why Uber tried to re-invent the wheel in some ways. Why not work WITH Volvo, considering they have an amazing track record for safety and their technology in their vehicles and had already developed a self-driving truck years ago.

[Gromitt]

Just to make it clear. The Volvo lorries in the video is not made by the same company that makes Volvo cars. They have nothing to do with each other. Volvo Cars is owned by the Chinese company Geely and Volvo Trucks is a company in Volvo Group.

Volvo Cars has not belonged to Volvo Group since 1999 when it was sold to Ford.

[nctnico]

It doesn't take a sophisticated AV to drive better than the average human driver.

I've heard this repeated like clockwork every time the safety issue comes up. But an AV system that cost over a billion dollars has killed a pedestrian after roughly 2 million vehicle-miles. The average human driver would have to drive half a billion miles to kill a pedestrian (12.5 fatalities per billion vehicle-miles in the US in 2016, of which 15% were pedestrians). All self-driving systems put together have only done 20 million miles or so, with multiple fatalities. Are you calling systems with many billions of dollars of development costs unsophisticated?

I think at this point it's safe to retire "AVs don't have to be very good to be better than humans" as a discredited argument.

Again: you have to seperate the distance travelled by the type of road to compare apples with apples because there are significant differences between the number of fatalities per km depending on the type of road. There is also insufficient data to determine the (statistical) distribution of fatal accidents. Besides that: where do you get to multiple fatalities? The Tesla autopilot crash doesn't count because it isn't an autonomous driving system. IOW: that someone got killed says zero about the chance it can happen again.

[MT]

I don't really see how you can draw any comparisons in the crash / mile driven rate between humans and AVs.

Are AVs driving in snow?  How about in pouring rain with high wind?  Dirt roads without lane markers?  AVs are only driving in ideal locations, under ideal weather, aka perfect conditions.  This isn't the case for human drivers that still need to get to work when its snowing.

These comparisons are rubbish.

From where do you draw hideous conclusion i made the comparison? Its Don MacKenzie , call him up and tell him
not me. You on the other hand should be upset that you have a uni lab that all day long deals with these issues
who then comes up with rubbish comparisons and get paid for it and referenced all over US Fake media! Are Uber
paying them? This uni have 7 nobel price winners!

[metrologist]

why would it be a rubbish comparison? I understand it's not valid, but does provide some value and insight into the program. The fact that AV are limited to ideal environments and have a significantly higher kill rate would seem relevant.

I think AV could work much better if they were limited to more heavily controlled environments - for example, shipping fleets that drive only along certain routes that provide adequate separation from pedestrians.

[Halcyon]

Just to make it clear. The Volvo lorries in the video is not made by the same company that makes Volvo cars. They have nothing to do with each other. Volvo Cars is owned by the Chinese company Geely and Volvo Trucks is a company in Volvo Group.

Volvo Cars has not belonged to Volvo Group since 1999 when it was sold to Ford.

I'm fully aware of this, and despite being independent entities, it would be naive to think that technology and design aren't being shared or co-developed. Volvo (as a whole) have a vested interest in building and maintaining the "safety" image in their vehicles. Whether it's driving one of their trucks or passenger cars, they are  designing their vehicles so that no pedestrian is killed by a Volvo in the future.

[Gromitt]

Just to make it clear. The Volvo lorries in the video is not made by the same company that makes Volvo cars. They have nothing to do with each other. Volvo Cars is owned by the Chinese company Geely and Volvo Trucks is a company in Volvo Group.

Volvo Cars has not belonged to Volvo Group since 1999 when it was sold to Ford.

I'm fully aware of this, and despite being independent entities, it would be naive to think that technology and design aren't being shared or co-developed. For example, the earlier Volvo S40 and the Ford Focus are essentially the same car.

Yes, but you put up a video about self drive Volvo lorries developed by Volvo Group and Volvo Cars does not belong to Volvo Group any more.

There are no co-development between Volvo Trucks and Volvo Cars, the only thing in common between them is the name/trademark Volvo, which Volvo Cars uses under licence from Volvo Group who ones it.

[orion242]

Shocked I say, shocked.  Someone massaging the numbers?

https://www.greencarreports.com/news/1107109_teslas-own-numbers-show-autopilot-has-higher-crash-rate-than-human-drivers

[JohnnyMalaria]

Shocked I say, shocked.  Someone massaging the numbers?

https://www.greencarreports.com/news/1107109_teslas-own-numbers-show-autopilot-has-higher-crash-rate-than-human-drivers

Wow. The author of that article clearly knows jack about statistics. Clearly nothing about the inherent foolhardiness of gross extrapolation. The arguments are just as vague and hand wavy as the data they try to poo-poo. The authoer doesn't even touch on the mentality of the people who can actually afford these cars and the immense false sense of security they get because they think it's driver-free. Don't believe me? See this act of stupidity and reckless endangerment of others:

Tesla driver who activated autopilot, hopped in passenger seat banned from driving

(Peeve alert: there's no such thing as a green car)

[orion242]

Clearly nothing about the inherent foolhardiness of gross extrapolation.

Yet we are constantly told they are safer, and who is putting these BS stats out there to begin with??

[RenThraysk]

Arizona Uber crash driver was 'watching TV'

The police report suggests the car's driver was streaming an episode of talent show The Voice rather than monitoring the car's progress. 

https://www.bbc.co.uk/news/technology-44574290

[Cyberdragon]

Arizona Uber crash driver was 'watching TV'

The police report suggests the car's driver was streaming an episode of talent show The Voice rather than monitoring the car's progress. 

https://www.bbc.co.uk/news/technology-44574290

Biased as hell. According to Dave's video, she was watching the console monitor.

Keep in mind it was dark and hard to see and according to Uber themselves, their system not only has NO EMERGENCY MANAGEMENT WHATSOEVER, no emergency brake, no emergency alarms...NOTHING!

That would be like a bar manager getting sued for using a third party robot to serve drinks and it starts throwing glasses at people.

"you should have been serving for it, it can't serve drinks!" (despite being advertised as so)

"What can it do then?"

"Uhhh...no comment!"

 

[maginnovision]

Arizona Uber crash driver was 'watching TV'

The police report suggests the car's driver was streaming an episode of talent show The Voice rather than monitoring the car's progress. 

https://www.bbc.co.uk/news/technology-44574290

Biased as hell. According to Dave's video, she was watching the console monitor.

Keep in mind it was dark and hard to see and according to Uber themselves, their system not only has NO EMERGENCY MANAGEMENT WHATSOEVER, no emergency brake, no emergency alarms...NOTHING!

That would be like a bar manager getting sued for using a third party robot to serve drinks and it starts throwing glasses at people.

"you should have been serving for it, it can't serve drinks!" (despite being advertised as so)

"What can it do then?"

"Uhhh...no comment!"

 

Ars technica also had an article on that watching hulu until the accident. Vehicular manslaughter is what this was and charges should be pressed. Possibly more since I'm pretty sure you're not supposed to watch tv on your phone while you're operating a vehicle.

[Cyberdragon]

If they gave her prior warnings that the car is not totally functional and has no emergency braking or alerts then it was her fault. If they misled her to believe the car would do anything at all if it sensed danger than it's totally Uber's fault for pedaling their stupid failure technology.

[Dr. Katzenfritz]

I develop automated driving functions for a living and have been extensively trained in driving this kind of vehicles. I find the whole media discussion about the case really bizarre, the weirdest was the US lawyer who claimed that here we'd seen the first case of the train car dilemma and that the vehicle would have clearly decided that it's the safest to run over the pedestrian.

The point is: The car was in development stage and thus (generally speaking, local law may vary slightly here and there) not a true autonomous car, but it was driven by the safety driver. Period. IIRC Google is now planning to ditch the safety driver, but as long as he/she's behind the wheel, there's no question about responsibility. Zero.
How the lady driving the vehicle could watch a TV show on her cellphone is totally beyond me. Our test vehicles are specifically designed to be operated without having to take the eyes off the road, like having single button measurement triggering and stuff like that. Even a short glance at the measurement screen is absolutely forbidden here, no one of my colleagues would even dream about taking out their cellphone...

Another point that the media misses in the whole debate:
One has to keep in mind that there is no big leap in the development of these systems which brings you from "normal car" to "Level 5" in one software release, but everything is sequentially tested and developed. The vast majority of test drives I did were with only very specific functions under test actually activated - in simplified terms, imagine you're working on keeping the lane, and while you're doing that you switch of any collision mitigation functions so they don't interfere with what you want to look at. That's not a safety issue at all because...*you* are driving the car!
In that context also, the media outrage that Uber deactivated the Volvo's own assistance systems is totally bewildering to me. Of course you do, the last thing you want is two different functions which don't know about each other interfering and sending the car off the road!

[maginnovision]

https://www.theverge.com/platform/amp/2019/11/19/20972584/uber-fault-self-driving-crash-ntsb-probable-cause

Uber was determined to be at fault. Finally.

[magic]

Get in front of a car expecting it to brake.
Die.

Imagine my shock. Some people have clearly never heard of things like drunk driving, texting while driving or plain old brake failure
I guess we can now add software bugs to the list

Uber was determined to be at fault. Finally.

Not exactly, and rightfully so.

Federal investigators split the blame for the fatal Uber self-driving crash between the ride-hailing company, the safety driver in the vehicle, the victim, and the state of Arizona

[SilverSolder]

Get in front of a car expecting it to brake.
[...]

A lot of drivers seem to take it for granted that the rest of the world will stop for them as they cheerfully poke their noses into the traffic without bothering to wait for a gap.

Pedestrians are generally more careful...

[coppice]

Get in front of a car expecting it to brake.
Die.

Only an idiot knowingly walks in front of fast moving traffic, but everyone does something that dumb from time to time. If one side of the interaction is alert there is a good chance of them saving the day. When one side is a dumb machine, the other side has little room for error. The reasonably low accident rate on modern roads has a lot to do with one side frequently taking evasive action. Its one of the reasons side impacts are such a problem - there is very little defensive or evasive behaviour you can use to mitigate them.

[nctnico]

Get in front of a car expecting it to brake.
[...]

A lot of drivers seem to take it for granted that the rest of the world will stop for them as they cheerfully poke their noses into the traffic without bothering to wait for a gap.

In Italy it is mandatory to drive like that unless you want to die of old age right there in your car. IOW: it greatly depends on the local driving style.