Future of humans driving motor vehicles.

kwackers
Posts: 15643
Joined: 4 Jun 2008, 9:29pm
Location: Warrington

Re: Future of humans driving motor vehicles.

Post by kwackers »

mattsccm wrote:Still wonder how this idea of driverless cars copes with non standard road/conditions.

Probably better than humans cope with standard road conditions I reckon.

I was sat in the right turn filter lane of a junction earlier behind two other cars waiting for oncoming traffic to stop and then we could turn.
I'd passed the stop line for the junction so not only was I blocking the lane to my right but I could no longer see any traffic lights.

Somewhat predictably the oncoming vehicles jumped the red light for as long as they thought was decent meaning the cross traffic now had a green light and of course that meant someone had to boot it from the light with their hand on the horn followed by traffic coming the other way also deciding that the proviso for a green light "if your way is clear" didn't apply to them either.

(I once saw a 4 car collision at the very same junction for the very same reason.)

It's not a difficult thing to get right and no doubt those same people who booted it from the lights without letting the junction clear or even considering what would happen if they didn't probably would have trouble understanding how a computer could drive a car when they themselves can't manage basic skills.

The answer as always is simply by avoiding getting yourself into such a position in the first place - something I suspect computers would be much better at.
User avatar
The utility cyclist
Posts: 3607
Joined: 22 Aug 2016, 12:28pm
Location: The first garden city

Re: Future of humans driving motor vehicles.

Post by The utility cyclist »

mattsccm wrote:Of course the concern about people being in control of vehicles is surely equally valid about cycles. Or legs for that matter. I still haven't found the piece of research that I came across some years ago showing that most pedestrian/vehicle accidents were the pedestrians fault. Eg idiots not looking when crossing road. By a huge proportion. Far fairer to ban walking I reckon. Still wonder how this idea of driverless cars copes with non standard road/conditions.
I would like to see the introduction, compulsory of technology that prevents vehicles exceeding speed limits. A sensor on the sign posts could do a lot. It could also completely blow up mobile phones used in cars and those nasty speed trap detectors.

Idiots not looking or people with a weapon not anticipating humans make small errors in judgement? The police are heinously complicit in the massive bias regarding 'at fault' when it comes to involvement with motorists and vulnerable road users and the way they record it for STATS19.

The group of road users that gets tanked from both ends are people on bikes, the ways the laws are applied both as a victim and alleged perpetrator are ridiculously slanted against them, you only need read/listen to how judges words in court for either end and the blame culture used by government and their various outlets to see how bad it is.
mattsccm
Posts: 5101
Joined: 28 Nov 2009, 9:44pm

Re: Future of humans driving motor vehicles.

Post by mattsccm »

Sorry wasn't clear 2nd time around. How do driverless vehicles cope with places that are not tarmac, standardised regarding markings, requiring give and take etc? Maybe squeezing through very tight gaps with some shunting. All/most of this would be day to day driving in rural areas. I have my doubts that it would work mores the pity.
mattsccm
Posts: 5101
Joined: 28 Nov 2009, 9:44pm

Re: Future of humans driving motor vehicles.

Post by mattsccm »

Idiots not looking. I watched someone do just that today. Was looking at his phone.
Vorpal
Moderator
Posts: 20700
Joined: 19 Jan 2009, 3:34pm
Location: Not there ;)

Re: Future of humans driving motor vehicles.

Post by Vorpal »

mattsccm wrote:Of course the concern about people being in control of vehicles is surely equally valid about cycles. Or legs for that matter.

People on bikes and legs aren't accompanied by 2 tonnes of steel & plastic to add to their momentum.

Unless they are filming the next Guardians of the Galaxy film, someone on legs is extremely unlikely to take out multiple other humans in one go.
“In some ways, it is easier to be a dissident, for then one is without responsibility.”
― Nelson Mandela, Long Walk to Freedom
Jdsk
Posts: 24635
Joined: 5 Mar 2019, 5:42pm

Re: Future of humans driving motor vehicles.

Post by Jdsk »

mattsccm wrote:Sorry wasn't clear 2nd time around. How do driverless vehicles cope with places that are not tarmac, standardised regarding markings, requiring give and take etc? Maybe squeezing through very tight gaps with some shunting. All/most of this would be day to day driving in rural areas. I have my doubts that it would work mores the pity.

Have a look at what Tesla Autopilot can already do.

And in general the answer is that they fail safe... give the problem back to the carbon-based lifeform... don't push on regardless... aren't thinking about something else or worrying about being late for a meeting...

Jonathan
kwackers
Posts: 15643
Joined: 4 Jun 2008, 9:29pm
Location: Warrington

Re: Future of humans driving motor vehicles.

Post by kwackers »

mattsccm wrote:Sorry wasn't clear 2nd time around. How do driverless vehicles cope with places that are not tarmac, standardised regarding markings, requiring give and take etc? Maybe squeezing through very tight gaps with some shunting. All/most of this would be day to day driving in rural areas. I have my doubts that it would work mores the pity.

If you look at some of the latest work in AI / ML it's pretty impressive.

What it's particularly good at is identifying 'stuff' which we think of as 'not well defined'. E.g spotting cancer on scans - which it routinely does more accurately now than specialists.
Identifying off road tracks is just more of the same, so providing it's suitably trained they shouldn't provide much of an issue for it.
Gaps and shunting is something a computer should potentially be pretty good at too although it depends how good the sensors are. In theory it should know to the millimetre how big the car and the gap is - something us humans aren't particularly good at. Even a good driver often gets it wrong if they don't leave suitably large gaps.

Of course offset against all this is cost.
Currently almost supercomputer power is required to perform these tasks and its not cheap. Tesla's for example have a stupid amount of computing power onboard.
On the other hand we're starting to see dedicated AI computer chips now so the cost is falling and the performance is going up. Bosch and folk like them are starting to make available 'off the shelf' hardware for a lot of the automation that's appearing.

I think that by the time this sort of tech becomes mainstream a lot of these points will probably be moot.
Jdsk
Posts: 24635
Joined: 5 Mar 2019, 5:42pm

Re: Future of humans driving motor vehicles.

Post by Jdsk »

kwackers wrote:E.g spotting cancer on scans - which it routinely does more accurately now than specialists.

I don't agree with that "routinely". There are some individual trials that show better performance in specific settings.

But I don't expect it to be long before that's widespread.

Jonathan
kwackers
Posts: 15643
Joined: 4 Jun 2008, 9:29pm
Location: Warrington

Re: Future of humans driving motor vehicles.

Post by kwackers »

Jdsk wrote:
kwackers wrote:E.g spotting cancer on scans - which it routinely does more accurately now than specialists.

I don't agree with that "routinely". There are some individual trials that show better performance in specific settings.

But I don't expect it to be long before that's widespread.

Jonathan

I guess you can argue around the details and yes most are trials - mainly because folk still don't trust the computers preferring the variability of humans.
I wonder what the difference between the best and worst humans are...

Anyway, like you said it won't be long.
I suspect it'll generally be a tool that specialists use before giving the scan the once over for that final approval.
Jdsk
Posts: 24635
Joined: 5 Mar 2019, 5:42pm

Re: Future of humans driving motor vehicles.

Post by Jdsk »

kwackers wrote:I guess you can argue around the details and yes most are trials - mainly because folk still don't trust the computers preferring the variability of humans.

We're currently so short of radiologists that resistance to change probably isn't too important. And BTW the outbreak has dramatically broken many other barriers in clinical practice.

kwackers wrote:I wonder what the difference between the best and worst humans are...

Very large. Removal of that inconsistency is an enormous driver to automation. More than a decade ago I had a conversation with a very senior health service manager. Computerised image recognition for cervical cytology screening had just become possible. We had big problems recruiting and retaining humans. I asked him if he'd trade a bit of performance for a lot of consistency. "Like a shot."

Spot the analogy with driving!

Jonathan
Vorpal
Moderator
Posts: 20700
Joined: 19 Jan 2009, 3:34pm
Location: Not there ;)

Re: Future of humans driving motor vehicles.

Post by Vorpal »

I really dislike the Jack Welch approach to managing human resources: keep firing the bottom 10% on the basis that you will constantly improve the pool of employees. The main problem with it is that it turns some folks into horrible backstabbers, who will do anything to keep their performance reviews in good shape, even if it is unethical or dishonest.

Driving and road safety management, however, is one of the few things I think could benefit from such an approach. If we gradually take away people's privelege to drive, starting with the worst, we can observe how the road environment improves, and (maybe) stop when it gets to an acceptable level. That will leave only the best drivers eligible to do it professionally.

Combined with traffic reduction measures, alternative delivery & travel modes, and the eventual replacement of human drivers with motor vehicle computers, we can protect vulnerable road users, and still provide ways for people and goods to get around.
“In some ways, it is easier to be a dissident, for then one is without responsibility.”
― Nelson Mandela, Long Walk to Freedom
kwackers
Posts: 15643
Joined: 4 Jun 2008, 9:29pm
Location: Warrington

Re: Future of humans driving motor vehicles.

Post by kwackers »

Vorpal wrote:I really dislike the Jack Welch approach to managing human resources: keep firing the bottom 10% on the basis that you will constantly improve the pool of employees. The main problem with it is that it turns some folks into horrible backstabbers, who will do anything to keep their performance reviews in good shape, even if it is unethical or dishonest.

Not only that but IME replacing them isn't that easy.
We interview people but in truth it tells you very little, some folk interview well but turn out to be crap, others interview badly but turn out to be a bit of a star (usually in other companies!)
With enough churn you may well eventually end up with a better than batting average of employees except the churn rate becomes so high and the atmosphere so toxic that you can't attract decent employees anyway.

IME a good mix of folk nearly always works better than cherry picking out the best. Although there's always a handful of people for which there's no hope.

(Probably very dependent on the industry I suspect though)
User avatar
6.5_lives_left
Posts: 112
Joined: 9 Oct 2020, 9:27pm

Re: Future of humans driving motor vehicles.

Post by 6.5_lives_left »

Not quite sure where to put this so resurrecting this thread.

I recently read an article on the Linux Weekly News site about a set of tools written by the Netherlands Forensic Institute to extract data from the computers on a Tesla Electric Vehicle and reconstruct what happened prior to the crash, all without having to rely on Tesla acting as a gatekeeper to tell the investigators what went wrong.

(Can you trust a company to be completely honest and not withhold information when the investigation outcome might have repercussions on the company and people working for or directing it?)

There is an aircraft industry investigation model where they recover the black box recorders and then all the interested parties are present while the recorders are examined.

The second link shows a video clip of a single vehicle car crash, presumably while the car was running on autopilot. Not very reassuring :shock:
Post Reply