Uber robot kills cyclist

brooksby wrote:

British jaywalking laws incoming in 1.. 2... 3... 

ArsTechnica has an interesting article on who's to blame for this incident: https://arstechnica.com/cars/2018/03/police-chief-uber-self-driving-car-...

ArsTechnica wrote:

Tempe police chief says victim "came from the shadows right into the roadway."

I don't know what technology Uber uses (probably the cheapest/worst judging by their ethos) but I thought that one advantage of driverless cars is being able to use Lidar type tech to be able to see things in the dark.

hawkinspeter wrote:

brooksby wrote:

British jaywalking laws incoming in 1.. 2... 3... 

ArsTechnica has an interesting article on who's to blame for this incident: https://arstechnica.com/cars/2018/03/police-chief-uber-self-driving-car-...

ArsTechnica wrote:

Tempe police chief says victim "came from the shadows right into the roadway."

I don't know what technology Uber uses (probably the cheapest/worst judging by their ethos) but I thought that one advantage of driverless cars is being able to use Lidar type tech to be able to see things in the dark.

The arstechnica article is interesting: seems like the usual arguments get passed out even when it's a robot driver. As you've said, the article starts to blame the victim because they came out of the shadows so the car didn't "see" her: so does that mean driverless cars will only be usable in daylight? Makes them a lot less attractive, I think...

hawkinspeter wrote:

Looks like the inevitable robot uprising has begun.

I, for one, welcome our new robot overlords.

'...once autonomous vehicles reach a critical mass they should have a calming effect on other traffic - forcing traffic to flow at or below posted speed limits.

Until they all get patched into SkyNet. Then we're screwed.

[/quote]

If the tech companies can make (flawed) self driving cars, why can't they make self limiting cars? Driver at the wheel, operating the controls and hopefully alert to hazards, but not able to break the speed limit, arrive at junctions at breakneck speed, go the wrong way down a motorway etc. etc. 

The know-how is clearly there, but not the political will.

Self driving cars seem to be the undisputable future, without much real debate or thought about the most appropriate application of the technology. 

Thoughts with the victims. A needless death.

pockstone wrote:

If the tech companies can make (flawed) self driving cars, why can't they make self limiting cars? Driver at the wheel, operating the controls and hopefully alert to hazards, but not able to break the speed limit, arrive at junctions at breakneck speed, go the wrong way down a motorway etc. etc. 

The know-how is clearly there, but not the political will.

Self driving cars seem to be the undisputable future, without much real debate or thought about the most appropriate application of the technology.

The problem with that idea is how do you market the cars to the poor drivers?

Approximately 90% of drivers think that they are better drivers than average, so you'd be left marketing this to the 10% of bad drivers who realise that they're bad drivers.

It'd be like selling a road bike with stabilisers fitted.

hawkinspeter wrote:

 

The problem with that idea is how do you market the cars to the poor drivers?

Don't 'market'. Enforce

It'd be like selling a road bike with stabilisers fitted.

Or brakes?

[/quote]

I see your point about the lure of the open road and the unfettered thrills sold by the car adverts, but something really has to be done to disabuse those poor drivers of their sense of entitlement to kill and injure. 

And if cars become crap to drive, maybe more people will cycle? (Careful what I wish for!)

pockstone wrote:

The problem with that idea is how do you market the cars to the poor drivers?

Don't 'market'. Enforce

It'd be like selling a road bike with stabilisers fitted.

Or brakes?

I see your point about the lure of the open road and the unfettered thrills sold by the car adverts, but something really has to be done to disabuse those poor drivers of their sense of entitlement to kill and injure. 

And if cars become crap to drive, maybe more people will cycle? (Careful what I wish for!)

There's a problem with the 'enforce' option - politicians have zero interest in penalising motorists, so we'd better appeal to market forces.

Stabilisers and brakes aren't the same - stabilisers help cyclists who can't balance whereas brakes are useful for novice and expert cyclists.

I don't think autonomous cars are going to be of much use until they get rid of the backup human driver requirement. Once they do, they can be marketed as allowing people to work/read/sleep whilst commuting (much like public transport).

[/quote]

'There's a problem with the 'enforce' option - politicians have zero interest in penalising motorists, so we'd better appeal to market forces. '

Enforce it  at the manufacturing/ certification stage. Incentivise it as well. Then the only drivers you need to worry about are those who disable or override the technology.

'Stabilisers and brakes aren't the same - stabilisers help cyclists who can't balance whereas brakes are useful for novice and expert cyclists.'

Brakes are (usually) considered an integral part of the bike's safety equipment . As would self limiting  technology if it were introduced. Perhaps it would be an' inconvenience' for  expert drivers, but one they'd have to accept for not  having to join the driverless revolution . If the technology is available, apply it where it might do most good , most quickly, rather than waiting for the nirvana of 100% driverless cars.

[/quote]  

hawkinspeter wrote:

pockstone wrote:

If the tech companies can make (flawed) self driving cars, why can't they make self limiting cars? Driver at the wheel, operating the controls and hopefully alert to hazards, but not able to break the speed limit, arrive at junctions at breakneck speed, go the wrong way down a motorway etc. etc. 

The know-how is clearly there, but not the political will.

Self driving cars seem to be the undisputable future, without much real debate or thought about the most appropriate application of the technology.

The problem with that idea is how do you market the cars to the poor drivers?

Approximately 90% of drivers think that they are better drivers than average, so you'd be left marketing this to the 10% of bad drivers who realise that they're bad drivers.

It'd be like selling a road bike with stabilisers fitted.

Nah - I reckon it'll be relatively easy for a goodly percentage, maybe even the vast majority. Whilst there will always be folk enjoy the interaction, or the routes they take, (and i'm one of those to some extent) I whole-heartedly believe that basically no-one really actually likes driving most, if not all, of their journies. It's nearly always shite in towns, where a lot of the journey instances actually are, and dull on the longer ones. Traffic jams and congestion require effort and input far beyond their benefit. Even simple navigation and route planning are things that many would happily turn over to automation. If you had something you could just jump into and go somewhere without having to actively control it, I think an awful lot of people would jump at the chance - part of the pleasure of train journies, of the convenience of taxis and the like. That is might actually reduce the horrific amount of incidents that result in injury and death, improve the flow of what (hopefully reducing) traffic there is and so reduce pollution, grid-lock, frustration and wasted time and hopefully start to turn the idea of 'car as prized asset' to more of 'car as service' (might be being a bit hopeful there...) is to be welcomed by all if it can be actually implemented safely.

At least that's the simple version.

CygnusX1]</p>

<p>[quote=ConcordeCX wrote:

hawkinspeter wrote:

 

The failure here wasn't the software in the car, it was the flawed "fail-safe" backup  behind the wheel. 

 

No such thing as a software failure - it's always a design failure.  In this case, the design clearly fails to resolve an appropriate solution to the problem presented.

nniff wrote:

No such thing as a software failure - it's always a design failure.

This isn't true at all. What if the software is not in accordance with the design?

nniff wrote:

No such thing as a software failure - it's always a design failure.  In this case, the design clearly fails to resolve an appropriate solution to the problem presented.

It is impressive that you have written millions of lines of code without a single error.

Are you an AI machine?

hawkinspeter wrote:

Looks like the inevitable robot uprising has begun.

This unfortunate incident highlights just how rubbish Level 2 autonomous driving is. It's unreasonable to expect a driver to be paying enough attention to suddenly take control when they're just sitting there without anything to focus on.

Autonomous vehicles only make much sense when they're level 3 or above.

I think level 4: remove the driver.

Now how do we get there, from here?

VeloUSA wrote:

Police determined it would be impossible for a human to see cyclist emerge from shadows to brake in time.

I'm very skeptical of this. The road was lined with street lights and Herzberg had already crossed the distance of a lane or so before she's visible on camera. She'd been in the road for some time.

Cameras exaggerate shadows because they don't have the dynamic range of the human eye. The video does not prove it would have been impossible to see her, just that the camera was not able to pick things up outside of the bright beam of the headlights. If visibility was truly as shown in the video, the car should have been going much slower. I suspect visibility is more accurately shown in the image from the sf chronicle article above.

This case is likely to set a precedent, i hope it will be investigated seriously.

CygnusX1 wrote:

So much for having a human operator behind the wheel to take over. RIP sister.

I've seen the video, and the driver was paying more attention to whatever system information the car was providing her than to the information out the windscreen. Definitely not how I would want a test to be run. Would want a driver to watch the road, and either a passenger looking at the data, or a check of the data afterwards. Uber cutting corners horribly.

DrJDog wrote:

CygnusX1 wrote:

So much for having a human operator behind the wheel to take over. RIP sister.

I've seen the video, and the driver was paying more attention to whatever system information the car was providing her than to the information out the windscreen. Definitely not how I would want a test to be run. Would want a driver to watch the road, and either a passenger looking at the data, or a check of the data afterwards. Uber cutting corners horribly.

I read that they say that they usually do have two humans on board, for precisely that reason. Why they didn't in this case, I can't say...?