Electric Self Drivng Cars vs Roads

[Richeh]

Bit of a weird one but thought I would ask your opinions.

I have always been of the mindset that self driving cars will never be fully acceptable unless roads are transitioned to 'Self Driving Roads'. Even today we see Tesla's etc randomly stopping or making driving mistakes when on Self Driving mode.

I genuinely think that unless all roads are upgraded with sensors or whatever they do etc self driving cars will never be fully accepted unless the roads support them?

This is of course a massive feat and I understand it would probably take at least 10+ years for UK roads to be adapted for this but does anyone else share the same opinion? Obviously American roads are much newer and it would be easier to implement this on a large scale but for the UK we basically made roads out of horse tracks etc so definitely longer for us.

I understand the positive side of Eletric vehicles but can see how the self driving option will always struggle unless the roads are upgraded and equipped and think self driving will never be able to be 'fully' accepted unless roads are upgraded etc

How do you feel about this point?

 

[Rroff]

Something I didn't realise until recently - most of the self driving models have huge issues with the more extremes of weather (along with stuff like complete inability to understand ice on roads) and are largely insurmountable issues as things stand.

 

[sovietspybob]

Full self driving is still a long way away, i also really don't think anyone will be upgrading roads to make them more compatible in an active sense. Who would pay for that? government? the auto manufacturers? local councils? I don't see any of them ever reaching into their pockets for something like this. If they needed to spend money on public infrastructure like that they'd be better served pouring that money into public transit and things like active transport like properly segregated bike lanes.

I think we'll see gradual improvements to basically posh lane keep assist we have today on lots of cars, they'll take the strain on motorways and bigger roads but i don't see much more than that for quite a long time, as Tesla has demonstrated even with billions of dollars and tens of thousands of hours of user data at their disposal it's still rubbish. The public don't seem to care about full self driving, it'd be more the likes of taxi firms and maybe longer distance deliveries that'd benefit so we're more likely to see it on trucks delivering between warehouses i think as the nitty gritty of getting something to reliably navigate a tiny lane covered in mud in the driving rain at night is going to be so hard there may not be a business case for it for years and years.

Also don't confuse self driving and electric cars, they're two distinct things. Yeah most of todays headlines about self driving are on electric models but the majority of electric cars are not self driving, they're just regular old cars which happen to have batteries instead of a petrol tank.

 

[Kenai]

The first step to acceptance of self driving vehicles is for people to agree on what frequency and severity of mistakes are acceptable in a self driving vehicle - they will never be perfect but how good is good enough? Humans are evidently nowhere near perfect and routinely make mistakes, have accidents and perhaps worse some people actively decide to do stupid things.

So where does the line get drawn to say self driving is good enough - As good as humans? Slighty better? Significantly better? Almost as good?

 

[Deleted User 298457]

Even today we see Tesla's etc randomly stopping or making driving mistakes when on Self Driving mode.

Citation needed. If you are talking about a handful of videos versus how many Tesla's are on the road, then this is misleading.

 

[Bear]

Citation needed. If you are talking about a handful of videos versus how many Tesla's are on the road, then this is misleading.

I presume he’s not talking about the normal autopilot people are using. In the states recently they trialled a downloaded version of full self driving and a self driven car decided to do an emergency stop on a live lane for no apparent reason and caused a pileup

edit: Heres the article https://theintercept.com/2023/01/10/tesla-crash-footage-autopilot/

 

[MikeTheNative]

Citation needed. If you are talking about a handful of videos versus how many Tesla's are on the road, then this is misleading.

Ars technica article about auto pilot.

NHTSA data shows Teslas using Autopilot crashed 273 times in less than a year

Honda had the second-highest number of crashes, with 90 over the same period.

arstechnica.com

**Tesla's Autopilot system has also been repeatedly singled out by the National Transportation Safety Board, which even blasted NHTSA for failing "to recognize the importance of ensuring that acceptable safeguards are in place so the vehicles do not operate outside of their operational design domains and beyond the capabilities of their system designs."**

 

[Jonnycoupe]

Motorways in dry weather and well painted roads is easy. Anything else = world of fun.

 

[Deleted User 298457]

I think it'd be naïve to think an average human is smarter than a self-driving car; especially because of the dodgy lanes/poor markings.

 

[jpaul]

Citation needed. If you are talking about a handful of videos versus how many Tesla's are on the road, then this is misleading.

starters - the tesla ACC sytem crashing into motorbikes because it's non-lidar system thinks their light spacing indicates bike is much further away - under legal accusation I believe.
newer/cheaper Lidar's systems may help watching CES announcements..

 

[Hotwired]

I think it'd be naïve to think an average human is smarter than a self-driving car; especially because of the dodgy lanes/poor markings.

It's a scam on the naïve to sell products with the tags "smart" and "AI" and get them believing there's anything more than a robot following a script.

Anything not on the script well that's too bad, car carries on at full speed or rolls the dice because scenario not recognised.

And who's responsible when the car is incapable of doing what an excellent driver can do? Oh well, lets blame the driver of the self driving car because the car is deliberately misnamed.

 

[Deleted User 298457]

It's a scam on the naïve to sell products with the tags "smart" and "AI" and get them believing there's anything more than a robot following a script.

Anything not on the script well that's too bad, car carries on at full speed or rolls the dice because scenario not recognised.

And who's responsible when the car is incapable of doing what an excellent driver can do? Oh well, lets blame the driver of the self driving car because the car is deliberately misnamed.

You can't be a technologist with a mind-set like this?

starters - the tesla ACC sytem crashing into motorbikes because it's non-lidar system thinks their light spacing indicates bike is much further away - under legal accusation I believe.
newer/cheaper Lidar's systems may help watching CES announcements..

Lidar will be a fail safe once the cars start communicating to each other.

 

[Hotwired]

You can't be a technologist with a mind-set like this?

Nonsense.

Recognising road markings is a tool not a final decision maker and that's your example for being "smart" at driving which is almost entirely about the decision making.

Any piece of junk can have some tools to control a vehicle and rudimentary "sight".

Where "is" the example of a really good decision making robot operating a car. Present tense was given by you which means it exists yes?

 

[MikeTheNative]

You can't be a technologist with a mind-set like this?


Lidar will be a fail safe once the cars start communicating to each other.

Tesla don't want any actual sensors on their cars. Luckily most other car makers disagree with this and are using them as part of FSD.

Former Head Of Tesla AI Explains Why They’ve Removed Sensors; Others Differ

Karpathy outlined the reasons for Tesla's controversial choices. Here's a discussion of those arguments and the counter-arguments within the industry

www.forbes.com

 

[MikeTheNative]

Nonsense.

Recognising road markings is a tool not a final decision maker and that's your example for being "smart" at driving which is almost entirely about the decision making.

Any piece of junk can have some tools to control a vehicle and rudimentary "sight".

Where "is" the example of a really good decision making robot operating a car. Present tense was given by you which means it exists yes?

There are some of those robo taxis in San Francisco ,I think, which have seem to be doing well. However the company behind them is being smart about the implementation and keeping them in the same operating area that is well know/trodden route by the AI. San Fran having decent weather most the the year probably helps them a lot as well.

 

[Mysterae_]

There's too many hurdles before this country is 100% self-driving and possibly improbable without major paradigm shifts in attitudes and behaviours.

100% means 0% humans driving which many including me would hate. However, that won't matter in 30 years or so when I'm no longer driving. The current and incoming crop of lazy ******** would love a car to drive them to their local latte veg bar when the time comes.

There's the ethics side - how does AI choose between running over a child or a 3 elderly persons when those are the only two choices?

Cars would become just plain utilitarian boxes so there will be no market competition, you probably won't even own one, just beckon one with an app.

What is actually the desired final result of self-driving cars? The ability to take the car to the pub with impunity or to reduce accidents? There's easier ways of fixing those problems.

 

[lordrobs]

I think we'll see gradual improvements to basically posh lane keep assist we have today on lots of cars, they'll take the strain on motorways and bigger roads but i don't see much more than that for quite a long time, as Tesla has demonstrated even with billions of dollars and tens of thousands of hours of user data at their disposal it's still rubbish.

This is how I see it TBH. I wouldn't be surprised for sections of motorway to be determined as suitable for self-drive (ie. the road markings are complete and to a set standard, the road surface doesn't replicate the surface of the moon, entrance/exit slips are to a certain standard etc.) where people can activate self drive to let the car cover a chunk of motorway miles under relatively standardised conditions in good weather. Outside of these zones, like you, I think full self drive is a LONG way off.

Lidar will be a fail safe once the cars start communicating to each other.

How can I get my bicycle communicating with one of Musk's 'S3XY' machines so it doesn't attempt to drive through me?

 

[Deleted User 298457]

How can I get my bicycle communicating with one of Musk's 'S3XY' machines so it doesn't attempt to drive through me?

Can't be that difficult, no? Mandate a tag. If you don't fit a tag then sucks 2bu if the 3 or 4 safety layers don't spot you.

 

[jpaul]

How can I get my bicycle communicating with one of Musk's 'S3XY' machines so it doesn't attempt to drive through me?

yes doesn't seem to have improved much, we will soon be extinct .. I don't know how tesla image recognition/response is for high-viz jackets

following bicycle registration plate thread, and AP respone to cyclists
these were interesting, albeit first with slightly older S radar

Tesla Model S owner tests Autopilot’s ability to avoid hitting bikers [Video]

 

[Greboth]

I think we'll see gradual improvements to basically posh lane keep assist we have today on lots of cars, they'll take the strain on motorways and bigger roads but i don't see much more than that for quite a long time

You bring up an important point which is "time". Given enough time, I think you'd have be naïve to think we won't end up with self driving cars doing every journey for us. The question is simply how long until it's a) possible and b) accepted. I think time also plays an important part in investment too - there's really only a handful of companies actually pursuing autonomy but as we get closer to it being a reality, more companies will get involved and progress will accelerate.

I think we are a long way off from wholly autonomous vehicles but probably not too far off from motorways and major A roads becoming autonomous. I love my cars and I think full autonomy would be a shame as there would be no joy in driving but there already is no joy in doing hours of motorway driving anyway so I'd welcome that change.