GarryP

So,

1. The car spotted the the pedestrian 6 seconds before impact, in plenty of time to brake.
2. The car could have braked.
3. Uber had disabled said braking, expecting the human driver to do so.
4. Human driver didn't.

Sounds to me like autonomous operation was better than the human, who ****ed up, and Uber were ****wits (which had been said before).

Beoz

[

Yep. That's pretty much the way it is in China. We agree on something.

Wol

That's not the way I read it: it appears to say that when the car was on full autonomy the drivers brake pedal was programmed to be inoperative - presumably because the system was so infallible.

GarryP

Nope, the human driver was expected to be doing emergency braking.

It wouldn't make much sense to have a human driver there for safety, and then not allow them to brake ....

This was a human driver, being inattentive, and not braking when they should - the whole automation thing is pretty much a red herring since it was basically out of the loop from a control perspective.

Shard

That's what I recall from the initial reports, human inatttention/error.

Once AV is deployed in a big way, there will probably be a micro industry of devices and reflectors that further support AV safety.

GarryP

It's kind of what I mean by things being instrumented up the wazzoo.

Today if there is an accident it tends to be six of one, half a dozen of the other, blame gets spread around indiscriminately.

In an automated world, there will be millisecond by millisecond data. If the cause is a human, or bad road design, or even a sign that hasn't been cleaned, the blame will be pointed away from the autonomous vehicle strongly, and with a pack of feral lawyers to back it up.

Which will make the roads no place for humans.

And I think the bicyclists will get it in the neck. They have generally tried to blame others for accidents they are involved in, but the reality is often different, and there is no way a company behind an autonomous vehicle is going to accept the blame. And cyclists have no insurance. It won't take many $50k judgements against them for cyclists to be forced from the roads.

Shard

That sounds a bit pessimistic on cyclists. If the AV's are driving in an orderly way (likely) then cycle accidents are likely to reduce. Compulsory insurance could cover any accidents that do happen.

Shard

The Bracknell engineers' estimate was about forty years too ambitious, but it's finally coming about...

Tokyo AV Taxi

Wol

You have the advantage, I haven't seen the report first hand, just this NS paragraph in the article. It's not very clear, but the way I read it was that the manual brakes were disconnected. No contest that the driver was not paying attention, but the fact remains that the lady was in full view of the car for quite a long time with no reaction from it.

GarryP

When you mentioned it, this is the link I found :

https://www.newscientist.com/article...-before-crash/

Shard

It does seem that the AI did not recognize the human. If I recall the video she suddenly appeared out of the darkness to meet her fate (not that darkness should inhibit the AI).

​​​​​​When AV are fully implemented there will still be unfortunate trolley-problem type situations (eg. pedestrian or driver) which will result in human deaths.

GarryP

Looks like the automation of jobs continue to pick up steam. 'Just walk out' stores seem to be hitting the early 'roll out' phase and given the vast number of low paid full and part time roles, it will hit the job market hard when the stores still left in the malls get rid of most of their staff (FFS how many nail bars can one mall have?)

https://medium.com/syncedreview/chec...al-d02a5af6938

Shard

At the moment the tech seems a bit awkward. Needing an "array of sensors and cameras". I think the model needs to be rethought. I read about a Chinese grocery where you go about the store selecting items on a device, whilst simultaneously the items are picked at the backend and then bagged for you by the time you get to the checkout. For a convenience store, a reworked vending machine model might work. I somehow don't have faith in the sensors system (costly to implement/maintain). Nevertheless, automation of retail jobs is inevitable.

GarryP

Look at it this way. If they can set up sensors and cameras such that they can recognise what items you pick up - then they can work out what items you look at but don't buy, those you buy without looking at the price, those items you forget and have to go back. They can also work out you didn't brush your hair, have two kids aged 7 and 3 who are badly behaved, are shopping just after the schools kick out, and followed that yummy mummy with the nice arse.

Or in short, the amount of information they can amass on you, in even a short timespan, is immense. So not only can they get rid of staff, they can start getting into really dynamic pricing - putting up the prices at certain time, dropping them at others, moving items around so that you actually buy more.

And when they really want to get rid of those expensive stores and move to online in a big way, they know EXACTLY what you buy, when, and can make it seamless.

No wonder they are all about the AI and the cameras - the data gathering they can sneak in for free is a large part of the point.

Amazulu

Please explain how a robot is going to do someone's nails any time soon?