Collection of Luddite opinions about level 5 autonomous vehicles :-)

[engle]

I normally ignore uninformed comments on public stock websites, but decided to reply to this one:

2019: "One of the main reasons it won't be successful is most riders will not get in one. I'm a Uber driver and I ask my riders if they would get in one and most of them said they will not get in a driverless car. And another reason it will fail is that having driverless cars is a bigger expense for Uber or Lyft. Right now the driver is responsible for the cars and the maintenance on the cars. But with driverless cars that expense transfers to Uber, and in turn will make the rider rates go up and they will lose a lot of business to traditional taxi's." (from MarketWatch)

My reply:

1905: "One of the main reasons horseless carriages will not be successful is most horse and buggy riders will not get in one. I'm a horse and buggy driver and I ask my riders if they would get in one and most of them say they will not get in a horseless carriage". Really?

From Horse and buggy - Wikipedia : "By the early 1910s, the number of automobiles had surpassed the number of buggies, but their use continued well into the 1920s in out of the way places".

My prediction: by the early 2030s, the number of fully autonomous electric vehicles will surpass the number of human-driven internal combustion engine vehicles, but their use will continue well into the 2040s in out of the way places".


Horse and buggy - Wikipedia

EN.WIKIPEDIA.ORG
"

 

[diplomat33]

I think the fear that some people have about autonomous cars stems from not knowing the technology. People are always afraid of the unknown and are more comfortable with what is familiar. So in the days of the horse and buggy, people were afraid of automobiles. In the early days of flight, people were afraid to fly. But eventually, as people become more familiar with the technology, that fear goes away. The same will be true with autonomous vehicles. Right now, autonomous cars are a big unknown to a lot of people so they are naturally a bit intimidated about riding in one: "Will this computer really be able to drive me safely? I don't know. Let me just stick with a normal car that I am familiar. I know how to drive a car." Once people see that the tech is there and computers really can drive a car safely, people will trust it. More than the technology aspect, Waymo is really helping with the PR of self-driving with their ride hailing service in Chandler, AZ. People are getting to experience first hand that an autonomous car can work.

I agree with your prediction that autonomous cars will eventually become the norm and non-autonomous cars will be relics of the past.

 

[electronblue]

In my view the question with autonomous cars is not one of acceptance, but one of succesful implementation. There are likely more technologies that showed great promise but failed to catch on than there are successful ones. In this case it has to work great.

For example there was a time when all this would probably have been said about private nuclear propulsion/power as well as private flight. Yet we are not flying to work in our nuclear-powered pods no matter how convenient that would be.

While a ”Level 5 autonomous car” seems like a no brainer, it is unknown whether anyone can implement it and/or whether some other type of technology will surpass it or replace the idea instead. (Maybe those personal quadcopters finally. Or transporters.)

Luckily for us Tesla will have Level 5 no geofence robotaxis out next year according to their Autonomy Day presentation so we’ll soon know.

 

[jboy210]

I wonder how many people getting on airliners know that in certain conditions, humans are not allowed to land the plane. The pilots cannot touch the controls until the wheels are firmly on the ground.

 

[kavyboy]

"...I ask my riders if they would get in one and most of them said they will not get in a driverless car. And another reason it will fail is that having driverless cars is a bigger expense for Uber or Lyft."

-- Uber Driver

"It is difficult to get a man to understand something when his salary depends upon his not understanding it."
-- Upton Sinclair

 

[electronblue]

I wonder how many people getting on airliners know that in certain conditions, humans are not allowed to land the plane. The pilots cannot touch the controls until the wheels are firmly on the ground.

I wonder how many people would get into Teslas if the cars applied the same rule in some instances to mandatory Autopilot (as it stands currently) control.

The quality requirements for that will be significant, just like they are in aviation.

 

[jboy210]

I wonder how many people would get into Teslas if the cars applied the same rule in some instances to mandatory Autopilot (as it stands currently) control.

The quality requirements for that will be significant, just like they are in aviation.

No doubt the current Autopilot version would not meet the necessary safety standard, but then neither could the first aviation autopilots. Tesla system are getting better, so Tesla, and their competitors, are getting close to meeting the safety standard for Level 4/5 every day.

 

[electronblue]

To be clear:

For me the question isn’t so much can someone reach Level 5 and will it change the world? I believe yes and yes. Doubting this could be construed as Ludditism, I guess, though of course we can’t be sure...

But the main question is: Can Tesla do it at the quality required? I would say doubting this would not be Ludditism, because Tesla’s track-record is spotty at best on this.

 

[MarkJ]

I've been thinking about true driverless cars, and I don't think it will ever (well, say in 30 years) happen and certainly not with Tesla's current sensors.
1) Even if driverless cars were statistically 10 x (or even 100 x) safer than driven cars there would always be an accident where the driverless car caused it where a driven car could have avoided it. Would you put your children in a driverless car knowing it might kill them without anyone making a mistake?
There are such a huge number of difficult corner cases to over come. Joining a fast moving road while turning across traffic in heavy rain?
Approaching a junction with worn out road markings and a low Sun.
People will take advantage of driverless cars, maybe even kids playing chicken with them.
But mainly, as soon as a tragic accident happens, where a driver could have avoided it, people will refuse to get in one.
Driverless cars will be the next big thing, that never happened.

 

[DukeofURL]

1) Even if driverless cars were statistically 10 x (or even 100 x) safer than driven cars there would always be an accident where the driverless car caused it where a driven car could have avoided it. Would you put your children in a driverless car knowing it might kill them without anyone making a mistake?

Would you put your children in a school bus, knowing nothing of the school bus driver's abilities and past performance? You just have to trust that they've gotten enough sleep, are attentive, and have an above-average skill level. These aren't even a worry with self-driving cars.

There are such a huge number of difficult corner cases to over come. Joining a fast moving road while turning across traffic in heavy rain?
Approaching a junction with worn out road markings and a low Sun.

These are dangerous driving conditions, the conditions won't be made any better or worse by changing the driver out, but if (as you said) the self driving cars are 10x better than humans, then these situations would be 10x less likely to result in an accident with a computer driver.

People will take advantage of driverless cars, maybe even kids playing chicken with them.

This is a wildly unsafe activity, and would result in the exact same outcome as if you did it today (likely severe injury or death for the car occupant when they swerve). Except it would be captured on video with a computerized car, and hopefully the perpetrators (or their parents) would be brought to justice.

But mainly, as soon as a tragic accident happens, where a driver could have avoided it, people will refuse to get in one.
Driverless cars will be the next big thing, that never happened.

This is going to be the toughest part. It's almost a certainty that self-driving cards will lower the incidence of accidents, and as such many less people will die. If the cars are, as you mentioned, 10x statistically less likely to get in an accident, that means instead of a sample set of 1000 people who die, only 100 will. It's a risk we'll need to accept. Are you willing to have 100 people die at the hands of an unsympathetic machine, so that 900 others may live? Or, are you willing to to allow 900 people to die who could have lived, in order to be able to place blame on a human for the deaths?

There will be many who call for the abolishment of self-driving cars, effectively asking for the level of fatalities to be kept up high, just so they can feel better about it. We have a very rough path ahead of us...

 

[MarkJ]

Hi,
Thanks for replying. I'm on a phone so quoting is a little tricky.
The thing is, those 900 people who didn't die because driverless cars are 10x safer are anonymous. Those 100 that die aren't.
The 10 x safer won't apply in all situations. Some scenarios, as I mentioned, will be much more difficult in a driverless car with Tesla's current sensors. Other situations will be safer.

 

[DukeofURL]

We really don't know what Tesla's full capabilities and limitations are at the moment. There is a lot of speculation out there based on current functionality, but just because you're eating the sausage doesn't mean you know how it's made, haha. In the AP investors' presentation they finally let on exactly why they think LiDAR isn't necessary (with 3D vision mapping), but it remains to be seen if that is in fact a winning strategy.

Personally, I'm excited to see what comes of this all. Realistically, I think AP3 is good enough for L3. Maybe if they're incredibly clever, they could push L4? I think L5 is still at least a couple hardware/software generations away though. I'm just speculating, but It would be nice to be proved wrong.

 

[diplomat33]

Personally, I'm excited to see what comes of this all. Realistically, I think AP3 is good enough for L3. Maybe if they're incredibly clever, they could push L4? I think L5 is still at least a couple hardware/software generations away though. I'm just speculating, but It would be nice to be proved wrong.

I'm excited too to see what comes next. I expect that once we start seeing the AP3 stuff come out, we will see some promising FSD stuff.

Vision is actually not the hardest part IMO. Tesla's has the cameras positioned around the car to see. The AP3 computer should be good enough to process all the data. And we know Tesla is grinding through the work of machine learning to develop the NN to recognize objects around the car. In time, Tesla will get that stuff done. The real tricky part will be what you do with that data. It's one thing to see the world around the car, the car still needs to make the right driving decisions based on what it sees. Making those right driving decisions in all the myriad of edge cases, is what is holding us back right now from deployable L4.

It is also worth noting that based on Elon's answer during Autonomy Day about deploying robotaxis in limited areas first, Tesla is aiming for L4 robotaxis first not L5 right away.

 

[drift]

I've been thinking about true driverless cars, and I don't think it will ever (well, say in 30 years) happen and certainly not with Tesla's current sensors.
1) Even if driverless cars were statistically 10 x (or even 100 x) safer than driven cars there would always be an accident where the driverless car caused it where a driven car could have avoided it. Would you put your children in a driverless car knowing it might kill them without anyone making a mistake?
There are such a huge number of difficult corner cases to over come. Joining a fast moving road while turning across traffic in heavy rain?
Approaching a junction with worn out road markings and a low Sun.
People will take advantage of driverless cars, maybe even kids playing chicken with them.
But mainly, as soon as a tragic accident happens, where a driver could have avoided it, people will refuse to get in one.
Driverle

I've been thinking about true driverless cars, and I don't think it will ever (well, say in 30 years) happen and certainly not with Tesla's current sensors.
1) Even if driverless cars were statistically 10 x (or even 100 x) safer than driven cars there would always be an accident where the driverless car caused it where a driven car could have avoided it. Would you put your children in a driverless car knowing it might kill them without anyone making a mistake?
There are such a huge number of difficult corner cases to over come. Joining a fast moving road while turning across traffic in heavy rain?
Approaching a junction with worn out road markings and a low Sun.
People will take advantage of driverless cars, maybe even kids playing chicken with them.
But mainly, as soon as a tragic accident happens, where a driver could have avoided it, people will refuse to get in one.
Driverless cars will be the next big thing, that never happened.

ss cars will be the next big thing, that never happened.

Everyday tens of thousands of accidents happen where a driver could have avoided it, but they don't.

 

[PhantomX]

There are a few situations that make me question whether L4 or L5 will exist with current level of technologies:

1. Costco parking lot in front of the entrance. There usually is no specific path for people to walk between parking lot and the entrance, and there are generally lots of cars trying to navigate the same space. How would an autonomous car handle that?

2. School zone: How would an autonomous car handle a crossing guard that can throw off the sequence of cars at intersection? And how would a L4/L5 car handle school pick up when the school has people guiding the car to spots (assigned at arrival) to pick up students?

3. A construction zone where the lane markers no longer exist and you have workers holding stop/go sign because both direction share one lane

4. A freeway construction that actually take the lane onto the other side of the freeway. Would the safety feature in the car freak out if it hits you are going against the traffic flow? (I have seen this on either the 60 or 10 freeway here in So. Cal. a few years ago)

 

[MarkJ]

4. A freeway construction that actually take the lane onto the other side of the freeway. Would the safety feature in the car freak out if it hits you are going against the traffic flow? (I have seen this on either the 60 or 10 freeway here in So. Cal. a few years ago)

Had this exact situation in the UK a few days ago (2019.28.3.1) and it handled it perfectly using NoA.

 

[tsh2]

Had this exact situation in the UK a few days ago (2019.28.3.1) and it handled it perfectly using NoA.

Worth noting that a 'contraflow' is a fairly common scenario in the UK (and Europe), although maybe not as common as it used to be. I was slightly surprised that I needed to explain the term to someone who is learning for the driving test recently.

 

[PhantomX]

Had this exact situation in the UK a few days ago (2019.28.3.1) and it handled it perfectly using NoA.

That's good to know! I have seen it on local street, but first time I had encountered on a freeway where they open up center divider and route part of traffic to the other side of the freeway.