Self Drive Error Mostly Down To Human Error? Right Oh!

[ToraToraTora]

https://www.bbc.co.uk/news/technology-50484172
Err what's wrong with that sentence? "self drive"?? how can that be human error? well unless it is referring to the whole mental idea in the first place.

[davebro]

It's to much to expect that a human will sit there studiously monitoring every second when the car supposedly "self-drives" 100% of the time! And then pick up on the 0.05% of times it doesn't. I'm sure I would be asleep.

[Old_Geezer]

You had one job...

[ToraToraTora]

you mean like a driver davebro?

[jno]

for test drives the humans are supposed to stay awake and not watch TV, so they can take over if anything goes wrong. This human was watching TV. The same would happen if the driver of a regular car decided to watch TV.

[davebro]

I would not take a job monitoring self-drive cars. I couldn't do it safely!

[ToraToraTora]

a driver of a regular car is driving it. What's kin point if the software is so tish it needs a human watching? The clue is in the name "self driving".

[jno]

it was still at the experimental stage. How are you supposed to try it out if you don't put a sentient human being in it as a fallback?

As we see, this one didn't have a sentient human being in it.

[ToraToraTora]

it should not be on the road until it's ready, ie never, someone dies because these numpties do not understand software.

[davebro]

The question is - the system must have been tested umpteen times for such a scenario on test tracks etc. so why did it fail in this instance?

[ToraToraTora]

because what they are trying to do is not possible. They have not realised it yet but they will. We cannot program a robot to make a cup of tea yet. This is like the wright brothers deciding they can land on the moon.

[jno]

you'd think so, davebro; but most software breaks down at one time or another. So you have a fallback in place ... or not. Regular cars are called in often enough because someone's found a fatal flaw.

[ToraToraTora]

software does not break down, it merely encounters situations that are not catered for.

[CrapAtCryptics]

There is no software that can account for the vagaries of human behaviour all of the time.

[ToraToraTora]

...correct, so it will never work.

[davebro]

I think their (an?) argument is that the software will fail less often than a human would so overall it is safer even if some "accidents" still happen.

[Old_Geezer]

It'll never be ready enough for some, but in the real world, extensive real world road tests are clearly necessary unless progress is to be deliberately stopped to the detriment of the human race.

[Old_Geezer]

The Wright Brothers decide they can't make it to the moon so stop work on any heavier than air flying contraption. Tells reporter, "We'd much prefer we lived in the dark ages. Whatever were we thinking of ?"

[mushroom25]

interesting quote I read recently by Christine Negroni, aviation writer, when asked "what's better - automation or human control?"

she replied that both were different. the automation is a rules-based system that will continue processing and computing while the house burns down around it. the human, on the other hand, will adapt as the situation changes.

[jno]

unfortunately the human in this situation did not adapt though, as OG says, she had one job...

[archibaldy]

It's coming and not quickly enough for me. Autonomous cars do not have to have a 100% safety record, they just need to perform better than humans and with 180,000 casualties and 1700 fatalities per year in the UK alone the bar really isn't that high.