Electric cars

[Tony.Williams]

I suspect that it will be a while before the programming of the computers in such vehicles will be good enough to be able to react appropriately to the huge range of possible situations - especially in anticipating situations which are just beginning to develop.

On the other hand, human drivers are far from error-free, so the accident rate overall might be lower. But I suspect that accidents caused by autonomous vehicles will attract much more negative publicity than those caused by careless drivers.

[Croydon13013]

I suspect that it will be a while before the programming of the computers in such vehicles will be good enough to be able to react appropriately to the huge range of possible situations - especially in anticipating situations which are just beginning to develop.

On the other hand, human drivers are far from error-free, so the accident rate overall might be lower. But I suspect that accidents caused by autonomous vehicles will attract much more negative publicity than those caused by careless drivers.

I suspect that the AIs in driverless cars are already better and safer than human drivers.

Final tweaks are being ironed out: http://www.bbc.co.uk/news/technology-40416606

[Tony.Williams]

Computers are excellent at doing what they are programmed to do, but it's the unexpected events which the programmers didn't consider that can cause problems. As with the Tesla car a few months ago, in which the system failed to recognise a white truck against a white sky, so the car slammed straight into it, killing the driver. No doubt the programmers will learn from each such event so it won't happen again, but I suspect there'll be a few more "learning experiences" occupying morgues before it's all sorted.

[Croydon13013]

Computers are excellent at doing what they are programmed to do, but it's the unexpected events which the programmers didn't consider that can cause problems. As with the Tesla car a few months ago, in which the system failed to recognise a white truck against a white sky, so the car slammed straight into it, killing the driver. No doubt the programmers will learn from each such event so it won't happen again, but I suspect there'll be a few more "learning experiences" occupying morgues before it's all sorted.

The Tesla fatality was the only one to have taken place so far in a self-driving car. Teslas alone have done more than 130 millions miles. The car average worldwide is one fatality for every 60 million miles.

The specific problem with not seeing the trailer (I think it was height off the ground rather than colour that really caused the problem) has been fixed. And worth noting that the driver had his hands on the steering wheel and could have overridden the autopilot. But he didn't see the trailer either.

Of course there will still be fatalities, and the scum press will misrepresent them. But we're already at the point where the computers are safer than human drivers.

[Tony.Williams]

The Tesla fatality was the only one to have taken place so far in a self-driving car. Teslas alone have done more than 130 millions miles. The car average worldwide is one fatality for every 60 million miles.

Is that for self-driving cars, self-driving Teslas or all Teslas? AFAIK the self-driving capability is a recent addition to Teslas, so probably hasn't seen a huge amount of use on the road.

The specific problem with not seeing the trailer (I think it was height off the ground rather than colour that really caused the problem) has been fixed. And worth noting that the driver had his hands on the steering wheel and could have overridden the autopilot. But he didn't see the trailer either.

See: http://uk.businessinsider.com/details-a ... ?r=US&IR=T

During the last 41 minutes of Brown's trip, the Model S was in Autopilot for 37.5 minutes. Brown had his hands off the wheel for a total of 37 minutes during the time the car was in Autopilot.

The Model S displayed the visual warning "hold steering wheel" seven times during the trip. Six of those were followed by auditory warnings.

An underlying problem is that when automated systems are installed in anything, people tend to rely on them and lose attention and even common sense. When sat-navs were first installed in cars, there were various stories of users driving into ridiculous situations because that's what the sat-nav told them to do.

[Ketchup]

I really don't think this 'driverless vehicle' thing will take off in a serious way. Great for a novelty for a while, yes, or contained within certain compartments of industry, but as for staying power for the masses .....

Thank goodness we humans haven't been programmed into lifeless, brainless robots - yet!

You Can Trick Self-Driving Cars by Defacing Street Signs -
https://www.bleepingcomputer.com/news/s ... eet-signs/

A team of eight researchers has discovered that by altering street signs, an adversary could confuse self-driving cars and cause their machine-learning systems to misclassify signs and take wrong decisions, potentially putting the lives of passengers in danger.

[Tony.Williams]

Something else I've read about the psychological aspects of self-driving cars: if the thing is running on automatic and the driver suddenly realises that he needs to take over for some reason, it can take several seconds for full manual control to be established - mainly because the driver isn't mentally ready for it.

[Matt]

Tesla's autopilot is not meant to be a self driving car. Driver is still meant to be in control.

[Tony.Williams]

Tesla's autopilot is not meant to be a self driving car. Driver is still meant to be in control.

That's the official line. But calling it an "autopilot" is just asking for trouble...

[Matt]

Tesla's autopilot is not meant to be a self driving car. Driver is still meant to be in control.

That's the official line. But calling it an "autopilot" is just asking for trouble...

[Tony.Williams]

Ah yes, I only saw that film once, decades ago, but I recall that scene: one of cinema's best laugh-out-loud moments!

[Tony.Williams]

I'd be curious to know the extent to which autonomous driving systems can anticipate problems. For instance, if we are driving along a suburban street, with cars parked on either side, and a ball suddenly bounces across the road in front, we automatically assume that a small child might well be in hot pursuit, so we slow down and prepare to brake violently if required. If, in addition, we see a child running in the direction of the street, we brake before the child even reaches it.

Does an autonomous system recognise a ball, and understand the implications? Can it take into account what is happening off the road? Or does it just react when someone suddenly appears right in front of it?

[Croydon13013]

Ah yes, I only saw that film once, decades ago, but I recall that scene: one of cinema's best laugh-out-loud moments!

Surely you're not serious?

[chaggle]

Ah yes, I only saw that film once, decades ago, but I recall that scene: one of cinema's best laugh-out-loud moments!

Surely you're not serious?

Don't...

[Tony.Williams]

Ah yes, I only saw that film once, decades ago, but I recall that scene: one of cinema's best laugh-out-loud moments!

Surely you're not serious?

Somewhat of an exaggeration, but I did laugh out loud...