Jaguar I-Pace EV400 Forum banner

1 - 20 of 32 Posts

·
Registered
Joined
·
980 Posts
Discussion Starter #1
I have always been skeptical of the idea that we can build fully automated driving cars (and I know quite a bit about the underlying technologies). I am not talking about driver assistance, lane hold, warning, auto brake, etc. I am talking about this over-hyped "auto pilot" idea. The main culprit pushing this lie is obviously Elon Musk. But my skepticism isn't just about Tesla, its an idea that captures everyones imagination but its simply not an option yet. It's just that Tesla pushing this narrative (just like his false assertions about hydrogen fuel cells) is counter productive.

This is an interesting article if you are interested.

https://arstechnica.com/cars/2019/03/dashcam-video-shows-tesla-steering-toward-lane-divider-again/
 

·
Registered
Joined
·
521 Posts
Tesla Announces It Is Not Testing Autonomous Cars In California

Tesla keeps autonomous driving test program opaque, says no testing on California public roads

eperali said:
I am talking about this over-hyped "auto pilot" idea. The main culprit pushing this lie is obviously Elon Musk.
I can agree that Musk is lying, and he may have been the earliest, but at this point Tesla is way behind GM Cruise and Waymo. Right now I see Mary Barra as the main pusher. She's been pushing her three zeros mantra for a while. She wants her robotaxi. It's seen as a potential money hose for GM.

They're pushing at NHTSA now:

GM's driverless car petition to NHTSA enters public comment phase

And they're ramping up further.

GM is doubling the staff for its self-driving car business

Their best so far is supercruise in Cadillacs which hasn't helped anyone commit techno-suicide yet AFAIK. Edmunds did a good comparison here.


Do I care? Yeah, if I have to share the road with them. I'll take the Supercruise. Tesla autopilot? No way. It nearly crashed the Edmunds guy. He saved himself. Thankfully he wasn't on his phone, or playing Jenga.


Tesla does not encourage what you see here. ^^^ But you could argue that they enable it.
 

·
Registered
Joined
·
980 Posts
Discussion Starter #3
Thanks that is part of my concern. The more conservative super cruise is more in line with reality and doesn't "encourage" bad behaviour, while Tesla's language, Musks hyperbole and what it "allows" does.
 

·
Registered
Joined
·
521 Posts
Right, but Supercruise doesn't work anywhere. It only works where Cadillac has geomapped the roads.

The biggest/best safety feature of Supercruise is that the wheel watches your eyes. Take up a game of Jenga, or hop into the passenger's seat and it shuts you down because it sees that you're not watching the road.

The Tesla will also stop the car if your hand isn't on the wheel periodically (you can argue too late to be of much help) but you can always use a water bottle to fool it.

 

·
Registered
Joined
·
521 Posts
Back to the original topic, I see a few things that will throttle it.

First is that LIDAR will still have issues with fog and snow, and a lack of road lines that are more common than anyone wants to admit. With geomapping they might get by the snow and lines problem, but in fog it's still semi-blind.

Second, there will probably never be a level 5 autonomous car. Mary Barra may be okay with that since the robotaxi can certainly be a money hose within city limits, at lower speeds. But it's unlikely to see highway speeds.

Third, machines will never be able to anticipate other drivers no matter how 'smart' they are. This is a matter of experience and learning that a machine can't and never will be capable of at the level humans are. That's not to say all humans are great at it, but the worst of us are still better than the best machine.

The autonomous car for the everyman won't happen in my lifetime. It's hard enough to afford an EV for most folks right now, much less a car with all that expensive tech attached. That's why Barra wants her robotaxi. It has to earn a regular income to be profitable.
 

·
Registered
Joined
·
112 Posts
I think I disagree here. A friend brought over his Tesla Model 3 today and we went out and played around with the ‘Autopilot by Navigation’ function and it is so good it is scary. Scary in a good way. While it had to ask your permission to change lanes (probably from a liability point of view) we drove for quite a while and it changed lanes, exited the highway and merged into traffic with no assistance from me. Yes you are correct—it isn’t available now, but man—it is so close I can taste it.

I also agree that it would be difficult from an infrastructure point of view, but what if they just geo-mapped or sensor embedded or painted lines with radio-active waste or whatever they come up with, just in those dense commuting corridors that suck out our collective will to live? What if they said by for instance, 2030, if you want to put your car on HWY ‘X’ it will have to have certain Level 5 sensors etc. I wouldn’t mind manually driving to the highway but then I would be perfectly happy to let the robots take over for the mind-numbing commute many of us do daily.

I for one eagerly await our new robot overlords.
 

·
Registered
Joined
·
521 Posts
Autonomous level 2 Camera fails, including at the site of the Mountainview death:


Another example of following a line to the left at 5 seconds in:


Not looking to keep debating Tesla. Frankly after 5+ years I'm sick of the subject. But these videos demonstrate how level 2 camera-based lane keeping can fail. The car did follow a line - the wrong one.

Before anyone suggests that cities and such should be made to keep the lines painted, I have this image to share.



Navigate that. And then there are these:

 

·
Registered
Joined
·
1,488 Posts
I like the level of Driver Assist features that the I-Pace offers. I am the primary driver, and the DA warns me if I'm making a mistake and keeps me centered in the lane. So I can look far ahead, and the car will keep me centered in the lane or warn me if I depart. It's very relaxing.

Yesterday, I saw people hitting their brakes about 1/4 mile ahead in the #1 lane of the freeway. So I knew something was up, most likely a stalled car or debris. I eased up on the following distance, hit my flashers and soon debris flew from under the car in front of me. A broken fiberglass ladder was in the lane. I went into the emergency lane and drove around it. It's not something that is that unusual. It happens perhaps every 3,000 miles of freeway driving. If I were letting the car act as primary driver, it lacks the ability to see up to 2 miles ahead and decode it. It's rare to go 100% on my brakes on the street.

I'm not sure how Cadillac, Tesla, Nissan, or Jaguar would handle that if left to it's own devices. But I predict it would fk up your car and possibly leave you stranded.

But I will admit: Everybody who uses a cellphone in traffic should have the car as the primary driver. It is amazing that we know distracted driving, especially cellphones, is increasing the accident rate, yet we as a species demand that our cars allow this behavior.
 

·
Registered
Joined
·
402 Posts
The thing that currently troubles me in my i-Pace (where I use the LKA & Radar Following regularly in Phoenix traffic) is the inability for the system to notice that the car in front of me has pulled into a turn lane (right or left) and is not in my lane anymore. When that car moves into a turn lane and slows down, suddenly my I-Pace slows down - hard and (to the person behind me) "totally randomly" as the lane in front of me is clear. I'm constantly overriding with my foot on the accelerator, but only after everybody in my car has been violently thrown forward against their seatbelts.
 

·
Registered
Joined
·
1,488 Posts
This will be a problem for Tesla. This is what the whole GM "Killer Key", "Exploding Trucks", and Ford's Pinto "Autoignition" issue was about in a legal sense.

While Urban Legends have corrupted otherwise intelligent folk, here's why you heard so much about those cases, and what the actual issue was:

All involved deaths. We have a hard time putting a value on human life. Even one death is too many according to lawyers, regardless of fault.

GM ignition keys did not have the lower rotational torque than many other cars. With a few keys on the rings, no impact on earth could rotate the keys. The problem occurred only when hanging a large mass off the key ring, which is true for all cars: Bad Idea. Damages lock, can cause unintentional rotation. Most the deaths were caused by rolling the cars, which turned off the ign, which by law disables the airbags.

So if it wasn't riskier than other cars, what was the problem? GM knew they lowered the rotational torque, eng'g found it acceptable. Then they tried to cover it up.

Saddle Tanks. Until NBC used rocket engines and a power drill to fake GM trucks catching fire (look at the fire, it starts before impact), GM fires were not more common than other brands, in fact, they were less common than Dodge. But GM had internal documents that investigated some of the fires, and decided to take no action since it was not a significant risk, nor worse than other brands. GM always contended this, and after the billion dollar settlement, an independent university study (Harvard? Carnegie-Mellon?) found GM was telling the truth.

So if it wasn't a real problem, why the settlement? GM knew the design could be improved to reduce fire risk even further, but did not do it.

And the Ford Pinto, the most famous of all Evil Automaker cases was no different. VW Beetles had a higher rate of fires by nearly 10 times. But Ford knew there was a fire risk putting the tank where all other American sedan makers were putting them, that is behind the axle. It was not significantly different in design than most cars sold.

So why was Ford singled out? They had a documented decision that chose not to improve further on an already low risk. But it never was more dangerous than cars of that era. Safer than some, more dangerous than others.

Now we have Tesla. Tesla knew that uncontrolled access highways could decapitate drivers because of a false sense of security by the driver. They did nothing to fix it. They made a financial decision like GM and Ford. They could have shut off the AP on uncontrolled access highways, or build in something in the software to recognize such threats specifically.

The one advantage that TSLA has, is a negative net worth. A $2 billion lawsuit won't fly. They would just file General Assignment or Bankruptcy to avoid it. Their pockets are not deep.
 

·
Registered
Joined
·
309 Posts
Third, machines will never be able to anticipate other drivers no matter how 'smart' they are. This is a matter of experience and learning that a machine can't and never will be capable of at the level humans are. That's not to say all humans are great at it, but the worst of us are still better than the best machine.
Humans are better today, but that won't last forever. Computers evolve faster than we do. I think Level 5 will happen, but it will take decades (or possibly some unexpected breakthrough a little sooner than that).

One million Tesla robotaxis next year? Elon Musk is a bold face liar.
 

·
Registered
Joined
·
980 Posts
Discussion Starter #16
Humans are better today, but that won't last forever. Computers evolve faster than we do. I think Level 5 will happen, but it will take decades (or possibly some unexpected breakthrough a little sooner than that).

One million Tesla robotaxis next year? Elon Musk is a bold face liar.
Agree mostly but my personal belief is unless every car is forced to be automated and only driven in controlled environment improvements in algorithms won’t become human intelligence, so this won’t be near. We may one day start developing true intelligence in automated systems but we not anywhere on that path, just better “algorithms.”

And you are right, Elon Musk is lying a lot, including the fact that he fears AI so his lies are worse than people realize.
 

·
Registered
Joined
·
898 Posts
I can't imagine the stupids of somebody who uses autopilot on a highway with traffic entering like that. I'd never do such a stupid thing. Then I can't imagine the stupids of somebody who knows that autopilot has issues in a certain situation on a certain road and they still use it there. Neither of these recent crashes has affected my overall positive view of autopilot. Used intelligently and as intended, autopilot is a great thing.

I use lane keep assist in the leftmost and second to leftmost line on a freeway without cross traffic bullshit. I keep my hands on the wheels and just manage to lower my stress level. I actually steer constantly, but it's like the system reads my mind and takes off much of the burden. Plus I don't usually have to break or accelerate. It's really nice.

And then when I want to get there quickly, I go into full madman mode with all the nannies off, and the car is a rocketship.

So I still am considering a Tesla. It's a real devil's bargain, though. Another ipace? Man... I want variety. If I can only sleep with one women for the rest of my life, at least let me drive multiple cars on a regular basis.
 

·
Registered
Joined
·
521 Posts
Humans are better today, but that won't last forever. Computers evolve faster than we do. I think Level 5 will happen, but it will take decades (or possibly some unexpected breakthrough a little sooner than that).
A computer that can anticipate human behavior? I'm having a little trouble with that concept. The algorithm for that would destroy itself. You're going to need a whole new type of processor for that, made of human brain cells. :wink2:

One million Tesla robotaxis next year? Elon Musk is a bold face liar.
See, I have this problem with liars. Once they lie, for whatever reason, I can't believe anything they say after that. Not even when they're so delusional that they actually believe their own BS. I have the same problem with Mary Barra and her robotaxi. However, at least Mary's car is proving itself on real city streets. I think that's going to be its limitation - nothing over 35-40 MPH (inner-city use only).
 

·
Registered
Joined
·
521 Posts
Agree mostly but my personal belief is unless every car is forced to be automated and only driven in controlled environment improvements in algorithms won’t become human intelligence, so this won’t be near.
You can have my steering wheel when you pry it from my cold dead hands. I don't like letting anyone else drive. But I'm going to be expected to trust a machine?


Maybe after they perfect secure foam.
 

·
Registered
Joined
·
309 Posts
A computer that can anticipate human behavior? I'm having a little trouble with that concept. The algorithm for that would destroy itself. You're going to need a whole new type of processor for that, made of human brain cells.
I have trouble with it too, but it's gonna happen. Really, it has already happened - it was 1997 when IBM's Deep Blue 'anticipated the human behavior' of Garry Kasparov, world champion of chess, and checkmated him. Google's DeepMind appears to be leaps and bounds beyond Deep Blue's cababilities, only 22 years later. Granted, driving (moreso than chess) is the sort of activity where humans tend to do better than computers. But really, it's just a matter of degree, and a matter of time. And when you get right down to it, how unpredictable is human behavior anyway? Perhaps less than you think...
 
1 - 20 of 32 Posts
Top