Maelochs

it sin't the cost, it is the fact that for every human who can look forward and see a better way, there are three or thirteen who claim that there is no improvement, never was and never will be.

Like that scientist said (paraphrasing,) new scientific ideas aren't "accepted," the people who believed the old ones die off.

noisebeam

Up next...

Self driving shopping carts on the MUP
https://www.azcentral.com/story/mone...law/315541002/

I-Like-To-Bike

Noisebeam, in your travels through your home area have you encountered many (or any) of the so-called driver-less vehicles successfully handling the difficult traffic scenarios you described recently, or any other situation which human drivers routinely successfully handle using their assertiveness, and flexible reasoning powers to know when the rules must be broken to proceed?

mr_bill

Uber statement on Uber Waymo settlement

-mr. bill

noisebeam

No. I benefit from their timidness.

mr_bill

Do Google cars follow all the laws....

-mr. bill

noisebeam

Can you use words?

mr_bill

-mr. bill

Ninety5rpm

Turns out "half way" was almost the end of the presentation - most of the 2nd half is Q&A. I finished this morning. Still wondering what you think I missed.

He confirms most of what I've been saying in this thread, including that these cars are not depending on V2V communications at all. He points out the chicken-egg problem with that approach. You can't deploy the cars until you have the infra; there is no point in developing the infra without the cars. Still, he sees room for V2V communications in the future, that's just not where there current focus is.

He also touches on how while they do obey the law, they teach the cars the exceptions too, like crossing the double yellow when necessary (which is strictly against the law in CA - no legal exceptions).

I like how they teach it to recognize anomalies and to automatically slow down or stop when they arise. The example of the salmon cyclist was very interesting, especially when he shared that he was convinced that had he been driving he would have hit the salmoning cyclist because his attention would have been consumed by the other cyclist.

noisebeam

Thanks. Yes I was aware of their handling of double yellow to pass cyclists as well. But it is the combination issues that I think are most challenging - for example a double yellow with a near steady stream of opposing traffic. Or making an illegal left because the thru lane is blocked with a crash (as I had to do cycling to work last week). No doubt there are many working on this and these near endless scenarios already.

Ninety5rpm

All I can say is that whatever rate you're encountering these situations, they're encountering them much more often because they are doing much more driving than you are, so they must be figuring out how to handle them.

Ninety5rpm

Read the other comments in this thread.

northernlights

Latest robocar accident: Tesla in autopilot mode rear ended a fire truck.

They can't even detect a huge object like a fire truck or semi, let alone a tiny bicycle.




Why Tesla's Autopilot Can't See A Stopped Fire Truck

https://www.wired.com/story/tesla-au...y-crash-radar/

Leisesturm

Coupla things. First, the driver could be lying. I mean... the car isn't going to argue about it. Second, the Tesla is not a self driving car. It is a semi-autonomous (driver assist) car. Ultimately the driver is still responsible for seeing the freaking fire engine (or the cyclist) and doing something about it. But you can bet your boots that Tesla is on the problem. That particular kind of rear end collision will not happen again in a way that exposes Tesla to criticism. I have a friend, who will remain nameless, s/he recently had an accident in which an expensive luxury car was totaled, because s/he was fishing around on the floor of the car for a dropped CD while in motion. I noticed that it was taking them a long time to get a new car and asked why. The answer was because they had two previous accidents already in whatever window the insurance company uses and would not be able to get any new car covered reasonably, so they were waiting for the window to time out on the oldest of the three accidents.

If AV's and semi-AV's prove to be as difficult to train out of bad driving habits as my friend, they will not flourish and will go the way of Chevrolet Corvairs. The marketplace is unforgiving. In the meantime I suggest we all prepare ourselves for a future with AV's in it. Our wanting them not to exist is not going to be enough to keep them from happening. Resistance is futile. I must once again highlight the incredible irony of the outpouring of hate and resentment towards autonomous cars when until fairly recently human drivers were the target of considerable fear and loathing in these threads. What happened? Was it all just a huge hoax? No matter. I for one can't wait till human drivers are all history. Good riddance.

Chris0516

I already read them. I referenced the attitude in aviation.

Because, Some airlines' push their pilots' to rely completely on automation. Some airlines push their pilots to rely completely on the pilot's own flying skills.

Then some airlines' push reliance on both automation, and flying skills.

The point I was trying to make, and the connection to aviation. Is that motorists' are taught how to drive a car. But they aren't taught how to use the ever-evolving automation, and there is no one, making sure they know it. Which gets to my original statement. The false-hood in the automotive industry is that automation will prevent accidents', and deaths. Not just of cyclists', and pedestrians, at slow speed. But also at high speed.

I-Like-To-Bike

Of course, wanting an event to happen, and spending lots of money on the process, is enough to assure it will "happen"!

Accordingly, for you and others of your ilk who wish that human drivers become history, any and all skepticism, doubt of viability, practicality, projected timelines for implementation, utility and/or alleged benefits of this alleged sure thing endeavor are nothing but the products of fear, resentment, loathing and hate from technical know-nothings. Uh huh got it!

Ninety5rpm

All that applies to concerns about driver-assist system, and I agree.

I'm a proponent of what Google, GM/Cruise and Aurora/VW are doing - true self-driving cars. No steering wheel. No driving skills necessary. Leave the license at home. Don't bother renewing.

I-Like-To-Bike

A "proponent"? That is putting it mildly, isn't it?

But weren't you a proponent of Uber and Tesla too just a month or two ago?

Ninety5rpm

I've been skeptical about what they were doing all along, but impressed they haven't had more problems despite their dubious approaches.

Chris Urmson discusses the difference in approaches in the lecture I keep asking you to watch. He says they put people in driver-assist cars and watched what they did (with cameras). That's what convinced them to go full self-driving. Not surprisingly, they very quickly learn to rely on the car to do the driving and fail to pay attention as they should. He said they even considered systems to shock drivers who looked away, but in the end decided to not even pursue driver assist. I think that's very wise.

On the other hand, Musk is getting tons of data from all his Tesla driver guinea pigs out there, who are crashing much less than I thought they would. That risky approach may pay off in the end.

northernlights

First of all, Tesla has admitted their autopilot technology cannot detect objects that are not moving. The article says Tesla acknowledges the fire truck incident was a result of that failure. The company does not deny the vehicle was in autopilot mode when it hit the fire truck, which proves the driver was telling the truth.



According to Waymo, that is a load of bull. It is not humanly possible to brake in time to avoid a collision at high speed when the vehicle fails to do so. Waymo states that anything below level 4 autonomy is inherently dangerous and Tesla autopilot is well below level 4.

Ninety5rpm

What's a load of bull per Waymo is that it's reasonable to expect a human driver to make up for driver assist failures. They're right about that.

But Testla's Autopilot is a driver assist system - it's not a fully autonomous vehicle. So this kind of crash does not reflect failure in autonomous vehicle technology.

Tesla's don't even have lidar sensors.

northernlights

Waymo never said that.


Waymo CEO John Krafcik explains danger of semi-autonomous systems - Business Insider

Waymo CEO John Krafcik explains why systems like Tesla's Autopilot could be a 'big problem'



What is the point of even having a self-driving car if you are forced to babysit it at all times because you can't trust it not to slam into the vehicle in front of you? At that point you're better off driving the car yourself.

Ninety5rpm

I think you must have misread my statement which I repeat, annotated:The red part is "a load of bull per Waymo".

In other words, they disagree that it's reasonable to expect a human driver to make up for driver assist.

They think it's unreasonable to expect a human driver to make up for driver assist.

They had their own driver assist system and they filmed what human drivers did with it. It was alarming. That's why they decided to go straight to fully autonomous.

I've been saying this for months.

Ninety5rpm

This is where this discussion went south, and why.

To what exactly in northernlights's post does your that refer?

1. the Tesla is not a self driving car.
2. It is a semi-autonomous (driver assist) car.
3. Ultimately the driver [in a semi-autonomous car] is still responsible for seeing the freaking fire engine (or the cyclist) and doing something about it.
4. All of the above?

I couldn't figure it out. I didn't think any of it was a load of bull accordingly to Google.

Yes, Tesla is not a self driving car. It is a semi-autonomous car. And drivers of semi-autonomous cars are still responsible for seeing the freaking fire engine.

None of that is a load of bull. Not per Google nor anyone else.

What was a load of bull, according to Google, and your quote of the CEO backs that up, was what I said: that it's reasonable to expect a human driver to make up for driver assist failures.

It's a load of bull because it's UNreasonable to expect a human driver to do that.

But our objection to your line of reasoning goes back to your earlier post where you referred to the failure of a semi-autonomous system as the "Latest robocar accident", as if that reflects on true autonomous car technology, like what Google, Cruise and Aurora are developing. It doesn't.

Maelochs

Which is


Not on point.

A Non-Self-Driving car hit a fire truck. Has ZERO to do with AI cars.

This incident with the Tesla has no more bearing on AI cars than do the accidents caused by drunk drivers.

There was no collisions between an AI car and a fire engine ... or anything else ... which has been mentioned here.