Self Driving Cars?

[NW-Bound]

The following video was recently posted by a Tesla 3 owner on July 15, 2019. He admitted to not paying attention when the "autopilot" was engaged. Be sure to read his comments explaining what happened, and his expectations about the system performance including the automatic braking system.

 

[NW-Bound]

And here's another one of Tesla 3 autopilot hiccup. This is posted recently on July 6, 2019.

In Vienna, a self-driving bus hit a pedestrian on July 19. The bus speed was limited to 7.5 mph, hence the caused injury was minor. The European city just started testing two small buses on June 6, and has halted the program.

See: https://europe.autonews.com/automakers/driverless-bus-hits-pedestrian-vienna.

 

[ERD50]

Wow. In the 1st, it seemed like the Tesla kept going fast for quite a while after hitting some cones?

The 2nd shows just how hard this is. With the curves, overpass, traffic moving on the lower level, that's a lot to process.

At some point, AI should be able to do this, just like AI can beat the great Jeopardy players. But it also depends on the sensors. Humans get so many cues in cases like that. Just like the Tesla that couldn't tell the broad side of a semi from background sky. Or the one that had to be told from a map that the upcoming overpass is not blocking the road, don't brake, it just looks that was because you are on a hill before that overpass, so the overpass appears to be at the level of the roadway. But a human sees this w/o a map.

Such a long way for this tech to go. Musk is promising all Teslas being robo-taxi capable in 2020? Oh boy.


Hey, I can't believe this thread has been dead for 6 months. I guess some of this self driving talk was going on in 'the other' (closed) thread?

-ERD50

 

[NW-Bound]

In the 1st video, I believe the driver woke up and stopped the car. Else, the car would keep on going and plow into the road service truck ahead, and the driver would have more than just a bad day.

In the 2nd video, as the poster/driver pointed out, visual cues that humans can recognize may be hard for a computer.

About Musk promising full self-driving capability, how many promises has he made on various subjects that have not panned out? I am betting he will never get it, not without all the sensors that Waymo puts on its cars, and Waymo does not say when it will be ready either.

About this thread being dead, have you noticed anything new on the media about this SDC technology? People forgot about this already, and they are paying attention to something else now, such as Prince Harry and his wife's daily life.

 

[explanade]

Tesla fans will tell you though that auto pilot is not self driving.

 

[NW-Bound]

^^^ It is certainly not. Ignorant people do not know the difference.

If people talk about some future systems with better capabilities, then we will know when we see how it works.

 

[JoeWras]

Hopefully we can continue to talk without going off the rails like the other thread did, because this is a good continuing topic.

Like Tesla or not, the two above clips show some interesting cases that many of us worry about. As long as Tesla owners are responsible, they are ultimately doing a service for the entire industry as owners run into some crazy cases. Wondering if the first one got confused by the "fake" oncoming traffic of the truck being towed. The second one is a vision problem. Humans can still look around that corner and see the stopped traffic. In time, car intelligence will catch up.

When this topic started, there were various camps as to how this will roll out. Some fast, some slow. It is becoming apparent to me that this will eventually just creep into our consciousness. We're going to have a hybrid period for quite a while as intelligent systems roll out. Tesla is pushing that curve. Other makers are being more conservative on their roll outs, but they are starting to introduce them.

Interesting times ahead. I hope we start seeing some reduction in fatalities as these features roll out. Unfortunately, I'm afraid many users will take the opportunity to text, sleep, have sex, etc. while the semi-autonomous systems operate. So, I'm not sure benefit will be seen right away.

 

[NW-Bound]

I suspect that mishaps or light accidents like the 1st video happen more often than reported in the media. Not everyone candidly went public like the driver/poster in the 1st video to warn people of the danger. For his admission of fault for relaxing his watch, he faced a lot of abuse from commenters (see his own note below the video). Even when the accidents resulted in a fender bender or even total loss of the car or injuries, they do not make the national news unless there is a fatality.

And then, I am sure there are a few orders of magnitudes more of the hiccups when the driver catches it in time to make corrective actions, as shown in the 2nd video. Again, we do not hear about this. Tesla owners probably have seen plenty, but many dismiss it as a mere inconvenience. And they claim that despite the flaws, it still helps reduce driver fatigue. This I cannot dispute nor agree to, because I do not have personal experience with a Tesla car.

Still, it is clearly demonstrated that Tesla's technology is far from being 100% self-driving to Level 4 or 5, in order to operate as a taxi on the city street. City driving is a lot harder than driving on a divided highway. And that's why other SDC developers have been spending the last few years concentrating on city driving.

 

[explanade]

Well Waymo never launched its robo taxi service in AZ. I think there’s a driver in those cars.

And this week, Cruise Automation, owned by GM, admitted that they won’t have SDC taxis in SF in 2019 as they previously announced.

 

[NW-Bound]

Waymo has a test program to haul some volunteered passengers around, for free I believe.

They claim the driver is there for safety and for the comfort of the passengers, and the car mostly drives itself. Waymo has not revealed how often the drivers have to intervene.

They are in town, and I see them every time I go out.

 

[Midpack]

Waymo has a test program to haul some volunteered passengers around, for free I believe.

They claim the driver is there for safety and for the comfort of the passengers, and the car mostly drives itself. Waymo has not revealed how often the drivers have to intervene.

They are in town, and I see them every time I go out.

While Waymo may not have revealed how often drivers have had to intervene for the current test program alone, they have revealed how often they had to intervene in all previous testing. It was posted earlier, and Waymo was far ahead of everyone else including Tesla. See graph below for example, the earlier data included Tesla (IIRC Teslas required intervention every 3 miles on average).

 

[NW-Bound]

...Waymo was far ahead of everyone else including Tesla...

I would expect Tesla to be at the bottom of the pack. I believe they pulled off the test program in California because that data had to be submitted to the state, and they did not look good.

If Tesla's system cannot even recognize the construction cones on the highway even now, how can one expect it to read stop lights, stop signs, see pedestrians, kids on bicycles, etc... ?

It is amazing how people like to listen to sales pitch, and want to believe in fairy tales.

 

[JoeWras]

For his admission of fault for relaxing his watch, he faced a lot of abuse from commenters (see his own note below the video).

I have a very small youtube presence of about 20 very short videos showing power tools in action.

I get all kinds of abuse. It is a sport for too many.

And that's the rub, anytime anybody tries to be helpful and put something on youtube, you will automatically become "the biggest idiot on earth", because that seems to be the easiest put down youtubers come up with. So, most people don't bother anymore. I've pretty much stopped, despite some very kind comments and thank yous too.

 

[Animorph]

Tesla isn't claiming that they currently have self driving, though some new car buyers certainly seem to believe that.

In the owner's manual Tesla says Autopilot is for divided highways only, should not be used in construction zones, and should not be used on surface streets. However, the car will give anything a try for the most part. It's not geo-fenced to prevent misuse. As long as the driver is paying attention it can be interesting to try some of the other uses. The car usually does pretty well. But it's not close to self driving.

Full Self Driving is supposed to be "feature complete" by the end of this year. It will have the much faster computer, operate with a much larger neural network, and should have more diverse training for the common self driving problems. It has little in common with the current software, which is close to maxing out the current computer's capability.

I assume that when Tesla FSD is first introduced this year (maybe) it will require plenty of supervision but might be able to make easy drives without having the driver take over. As it gets used Tesla will be monitoring FSD disconnects and feeding that data back into the neural network training. We'll all be test driving for Tesla. Then Tesla will iterate the software and improve its performance. That will never end, and who knows when the performance will be better than a human driver, or approved for official no-driver use. Or if Tesla might need another hardware iteration.

Until then, it's fun being a test driver. It is nice not to have to make all those little steering and speed corrections on long trips. But my hand is on the wheel and I'm looking down the road for problems.

 

[NW-Bound]

I don't follow Tesla news that close, yet ran across the announcement of "new and improved" hardware for more capabilities. It should be better than the current production system, but will it be good enough? I have seen Musk make all kind of promises before, so time will tell.

I assume that when Tesla FSD is first introduced this year (maybe) it will require plenty of supervision but might be able to make easy drives without having the driver take over. As it gets used Tesla will be monitoring FSD disconnects and feeding that data back into the neural network training. We'll all be test driving for Tesla. Then Tesla will iterate the software and improve its performance. That will never end, and who knows when the performance will be better than a human driver, or approved for official no-driver use. Or if Tesla might need another hardware iteration...

About using the users' real-life experience to train the entire fleet, I thought Musk promised that long ago. Yet, a particular car driven over the same street over and over never learned anything itself, let alone sharing anything with any other car.

The guy who killed himself when running into the road barrier in Mountain View, CA, drove over the same freeway every day for his daily commute. Maybe he thought that the car learned over time, and trusted it that final time. Yet, he was a software engineer for Apple. He trusted the system.

Maybe it will be different this time.

 

[NW-Bound]

I have a very small youtube presence of about 20 very short videos showing power tools in action.

I get all kinds of abuse. It is a sport for too many...

I usually do not read comments on Youtube videos. Many are made by juveniles. Enough said.

 

[NW-Bound]

Waymo has progressed to testing its cars in less than perfect weather conditions. They are testing in dust storms and foggy conditions, rain, and snow. They also have a wind tunnel for testing.

See: https://www.msn.com/en-us/autos/new...e-navigate-a-dust-storm-in-arizona/ar-AAGjjVh.

 

[rk911]

i don't have to pay attention cuz the car will save me. i know it's gonna happen. not too soon, i hope.

 

[Midpack]

Waymo = Under promise, over deliver.
[-]Tesla[/-] Elon Musk = Over promise, under deliver.

As brilliant as Elon is and as remarkable as Tesla is, he repeatedly gets out over his skis with the claims he makes re: “self driving” and Tesla has to backtrack - it’s been we’ll documented over the past 4 years. I hope Tesla gets there, but it would be nice if Elon would quit making bold premature claims, he continues to add to the confusion.

Even Tesla’s public statements about Autopilot contradict Musk’s comments. When there’s been a crash involving Autopilot (and there have been a few), Tesla will inevitably release a statement hammering this point: drivers need to pay attention to the road, and Autopilot does not prevent all accidents.

”Full self-driving” has been a long time coming. Four years ago, Musk claimed that Tesla’s vehicles would be capable of completely driving themselves without any human interaction by 2017. Two years ago, Musk announced every car made going forward would have the hardware necessary to facilitate this goal. Tesla has spent the years since advertising this impending breakthrough on its website as an easy add-on to the purchase of a new car, something that only required a few thousand dollars and a little bit of patience.

This isn’t the first time Musk has sparked controversy about Autopilot. Last April, Musk took his hands off the wheel after activating Autopilot in a Model 3 during an interview with CBS This Morning. He did it again during a 60 Minutes interview in December. It’s like he can hardly help himself.

https://www.theverge.com/2019/1/30/18204427/tesla-autopilot-elon-musk-full-self-driving-confusion

 

[JoeWras]

As brilliant as Elon is and as remarkable as Tesla is, he repeatedly gets out over his skis with the claims he makes re: “self driving” and Tesla has to backtrack - it’s been we’ll documented over the past 4 years. I hope Tesla gets there, but it would be nice if Elon would quit making bold premature claims, he continues to add to the confusion.

+1

 

[kgtest]

+1

Traded my 3/4 ton for my buddies Dual Motor 3 yesterday. After a day of cruising in the Tesla, I am happy to say it's not for me.

The technology is certainly a bit above their competitors, but I'm happy to have my 3/4 ton back.

Something about driving a tank around that gives me peace of mind...

Curb Weight: Dual Motor - 4,072 lb
Curb Weight: Diesel Duranamax - 7,620 lb

I will say this, the acceleration I experienced was fun in the Tesla, the interior was soo cheaply made/looking that it fell out of my taste. I'd much rather take the Cadillac Super Cruise in terms of interiors.

I've driven a Audi RS, Viper R/T 10, older fully blown overhauled 5.0 mustang (the old square model), and some crotch rockets etc. Similar thrill with the Tesla. BUT if you have never driven ANYOTHER performance car, and Tesla was your first experience with fast acceleration, I could see how you could be "blown away" by that. For me, it just felt like another performance car.

I'm gonna rent a Stingray in West Palm Beach in October, I'll let you know how that compares.

 

[NW-Bound]

Waymo = Under promise, over deliver.
[-]Tesla[/-] Elon Musk = Over promise, under deliver.

Musk had to promise a lot of great things about Tesla autopilot, because it's one of the things to sell his cars. He is cash strapped and desperate. Google on the other hand is sitting on a lot of cash, and does not need to make immediate money on this technology.

Look at how much money Google is spending on testing, while Tesla is using car owners as test guinea pigs, and sells that feature as a plus.

Tesla has some very good engineers, but when you do not have enough money nor time, and have a boss breathing over your neck, it is that much harder.

 

[davebarnes]

You people need to be patient.
2040 - autonomous vehicles are here and work well
2060 - human drivers are illegal (in most cases)
2070 - mass transit systems in the USA collapse

 

[Midpack]

You people need to be patient.
2040 - autonomous vehicles are here and work well
2060 - human drivers are illegal (in most cases)
2070 - mass transit systems in the USA collapse

When does your crystal ball say Level 4 or 5 SDC’s reach 50% market share? The transition is going to be a much tougher and longer period than whenever SDCs become commonplace. Predicting 2040 is pretty safe since the transition may take decades.

 

[MichaelB]

When does your crystal ball say Level 4 or 5 SDC’s reach 50% market share?

In the year 2525?