Elon Musk is a nice chap

[sandalcandal]

no ego

Did you somehow overlook the title of the thread and "Elon Musk"?

Did you some how overlook every other word of CatalinaWOW's post? "no ego" is talking about "automated driving systems" not "Elon Musk" unless you are somehow suggesting autonomous driving systems are somehow magically imprinted with the human personality traits of a CEO ... hang on ... I think I hear the whisper of Steve Job's ghost from my iPhone

[bdunham7]

My only disagreement comes because I suspect that few drivers of non-automated cars are 100% effective.  They fail on edge cases too, and also mainstream cases because of boredom, distraction , medical issues and a host of other problems.  In an ideal world the automated driving system doesn't have to be perfect just better than human drivers.

I don't think that automated cars will be judged against human drivers and deemed satisfactory just because they nominally exceed humans on a statistical basis.  They probably will have to be at least an order of magnitude better than humans and not make any of certain classes of errors in order to be widely accepted.  A large portion of people are not going to accept automated systems that drive right into stopped trucks or stationary highway dividers.  Those that drink the kool-aid and do accept this or accept the excuses for it are in the minority, although that minority may be enough of a market for now.

[bdunham7]

unless you are somehow suggesting autonomous driving systems are somehow magically imprinted with the human personality traits of a CEO

In a way I think they are!

[EEVblog]

I don't think that automated cars will be judged against human drivers and deemed satisfactory just because they nominally exceed humans on a statistical basis.  They probably will have to be at least an order of magnitude better than humans and not make any of certain classes of errors in order to be widely accepted. 

Yep.

[EEVblog]

I'm pretty sure this point is known and acknowledged as a challenge. 10:27, 13:09. Sandy is blaming poor road markings for difficulties but Elon is asserting that an autonomous driving system must remain safe under any condition whatsoever.

Elon is slowly coming to realise this, and it's why he recently tweeted that, essentially, true autonomous cars will require a dramatic leap in AI (that we don't currently posses).

[CatalinaWOW]

I don't think that automated cars will be judged against human drivers and deemed satisfactory just because they nominally exceed humans on a statistical basis.  They probably will have to be at least an order of magnitude better than humans and not make any of certain classes of errors in order to be widely accepted. 

Yep.

You note that I said "in an ideal world".  In the real world an automated system drives into to the back of a truck and it makes world news.  I personally know two human drivers who have done it and it didn't even make local news (both survived, but long hospital stays and months to a not quite complete recovery)..  The total number of human drivers who have performed this feat has to be in the thousands.

I actually believe it is the lawyers who will end up deciding this, but can't predict which group of lawyers will win.  The ones sueing car companies for fielding a less than perfect autonomous system, or the lawyers sueing car companies for not fielding an autonomous system that would have prevented a stupid driver death.

[EEVblog]

I don't think that automated cars will be judged against human drivers and deemed satisfactory just because they nominally exceed humans on a statistical basis.  They probably will have to be at least an order of magnitude better than humans and not make any of certain classes of errors in order to be widely accepted. 

Yep.

You note that I said "in an ideal world".  In the real world an automated system drives into to the back of a truck and it makes world news.  I personally know two human drivers who have done it and it didn't even make local news (both survived, but long hospital stays and months to a not quite complete recovery)..  The total number of human drivers who have performed this feat has to be in the thousands.

I actually believe it is the lawyers who will end up deciding this, but can't predict which group of lawyers will win.  The ones sueing car companies for fielding a less than perfect autonomous system, or the lawyers sueing car companies for not fielding an autonomous system that would have prevented a stupid driver death.

Nope. Public perception is what will decide, like it or not, statistics based or not.
But yes, legalities will play a big role, and can indeed stop the introduction of them. But putting that aside, if fully autonymous cars are possible, people have to decide they want them.

[tszaboo]

I'm pretty sure this point is known and acknowledged as a challenge. 10:27, 13:09. Sandy is blaming poor road markings for difficulties but Elon is asserting that an autonomous driving system must remain safe under any condition whatsoever.

Elon is slowly coming to realise this, and it's why he recently tweeted that, essentially, true autonomous cars will require a dramatic leap in AI (that we don't currently posses).

Oh, and here I was and, I thought was only question of a "little bit of more software", right?
I wonder if Tesla owners can bring him to court and admit that all the cars sold with the Full self driving package that could one day drive themselves, will never be able to do just that. False advertising. Costs 7500 EUR, how many of this were sold?

[Black Phoenix]

Oh, and here I was and, I thought was only question of a "little bit of more software", right?
I wonder if Tesla owners can bring him to court and admit that all the cars sold with the Full self driving package that could one day drive themselves, will never be able to do just that. False advertising. Costs 7500 EUR, how many of this were sold?

I'm not saying anything that is mind blow and not obvious but autonomous cars will REALLY only happen when:

• The human is totally removed from equation, in a way to take over the driving of the car or being able to open a window or a door when the car is moving (people are stupid);
• Roads are totally enclosed to any non autonomous mean of transportation, protected from any external interference (humans, animals, debris, etc);
• Both the autonomous vehicle as the road are littered with cameras, LADAR systems, infrared, microwaves or any kind of precise position assessment of anything in the range of millimetres or even less;
• All elements on the road communicate between themselves to a central entity that decides the best course of action for any and every situation, while commanding all the elements in a "synchronous dance" kinda like "Bees in a Hive", without any way of intervention of a human.

When all of that is present and working then you will have autonomous vehicles and a reduction if not even a total eradication of accidents. And for me that will only happen 50 years from now.

What we have now is simply a great conversation starter if you are bored but not more than that. Is something more that it was 10 years ago yes, but still is something fancy for you to show off and nothing more.

[nctnico]

I'm pretty sure this point is known and acknowledged as a challenge. 10:27, 13:09. Sandy is blaming poor road markings for difficulties but Elon is asserting that an autonomous driving system must remain safe under any condition whatsoever.

Elon is slowly coming to realise this, and it's why he recently tweeted that, essentially, true autonomous cars will require a dramatic leap in AI (that we don't currently posses).

IMHO it is the wrong way to go to want a self driving car to work in every condition and/or to depend fully on AI; that is a fool's quest. The roads will need to be adapted to support autonomous vehicles in order to be better suited for those vehicles. Start with full autonomic driving on highways & main roads in/outside cities and use the system as an anti-collission / smart cruise control aid in non-autonomous driving areas. In Germany you can find these signs along some roads:

These are placed to let autonomous vehicles know their position. In addition to that they are also building a digital infrastructure to support vehicle to vehicle communication.

[sandalcandal]

IMHO it is the wrong way to go to want a self driving car to work in every condition and/or to depend fully on AI; that is a fool's quest. The roads will need to be adapted to support autonomous vehicles in order to be better suited for those vehicles. Start with full autonomic driving on highways & main roads in/outside cities and use the system as an anti-collission / smart cruise control aid in non-autonomous driving areas. In Germany you can find these signs along some roads:
[img]
These are placed to let autonomous vehicles know their position. In addition to that they are also building a digital infrastructure to support vehicle to vehicle communication.

My understanding of the consensus against external "marker" (visual, RF or otherwise) based self-driving is that the potential for interference (intentional or not) is too high. If external infrastructure based systems are required for the self-driving system to operate safely then the system will not be safe. Having a vehicle stop if markers are lost is obviously also dangerous and impractical. This is mostly what I heard second hand from my friend that does research in such systems.

Also LIDAR and radar systems are allegedly too prone to interference in real-life practical conditions. Apparently "phantom braking" which is an issue with driving safety/cruise control systems from multiple automakers https://www.baileyglasser.com/news-lawsuit-targets-nissan-and-infiniti-vehicles-for-faulty-phantom-braking is caused by radar interference or false signals. This is the reason I see cited for choosing MV as a more robust means of achieving safe self-driving.
https://www.voltequity.com/article/why-lidar-is-doomed (tried to find one without a Tesla slant but I guess Tesla is too prevalent )

Not an expert here so open to alternative points.

Also people seem to feel there's been a shifting stance from Tesla/Elon Musk on attitudes towards self-driving but I haven't seen it so I don't know where that's coming from.

[nctnico]

IMHO it is the wrong way to go to want a self driving car to work in every condition and/or to depend fully on AI; that is a fool's quest. The roads will need to be adapted to support autonomous vehicles in order to be better suited for those vehicles. Start with full autonomic driving on highways & main roads in/outside cities and use the system as an anti-collission / smart cruise control aid in non-autonomous driving areas. In Germany you can find these signs along some roads:
[img]
These are placed to let autonomous vehicles know their position. In addition to that they are also building a digital infrastructure to support vehicle to vehicle communication.

My understanding of the consensus against external "marker" (visual, RF or otherwise) based self-driving is that the potential for interference (intentional or not) is too high. If external infrastructure based systems are required for the self-driving system to operate safely then the system will not be safe. Having a vehicle stop if markers are lost is obviously also dangerous and impractical.

You have to ask yourself: how is that any different than driving through sudden thick fog or heavy rain? In the end there is only so much you can do and piling on edge cases is going nowhere. If I look at the accidents that have happened with Tesla's FSD then I want to seperate these in 3 categories: ones that are unavoidable either way, ones that happened in places which are unsuitable for self driving cars and ones that could have been avoided by adapting the road. The latter case is where the low hanging fruit is and where you can get the biggest benefit in terms of increasing the capacity of roads / reducing congestion.

[EEVblog]

If I look at the accidents that have happened with Tesla's FSD then I want to seperate these in 3 categories: ones that are unavoidable either way, ones that happened in places which are unsuitable for self driving cars and ones that could have been avoided by adapting the road. The latter case is where the low hanging fruit is and where you can get the biggest benefit in terms of increasing the capacity of roads / reducing congestion.

Remember, Tesla FSD is called Full Self Driving, but that's in name only. It's not actually full self driving. It's actual name is AutoPilot, and that's what it officially is, and advanced autopilot that can lane change and brake etc. And it has a cute auto-summon mode where it find you in a carpark or something. They are very clear about this on their website. It's Musk that has been personally talking up "Full Self Driving" and fully autonymos for years now, or at least implying it. There are countless FSD beta video out there showing how dumb the system is in even basic almost ideal condition scenarios.
He's either had a technical awakening or the suits have tapped him on the shoulder to calm it down, they aren't even remotely close to full autonymous self self driving which Musk now has essentially admitted.

https://www.theverge.com/2021/5/7/22424592/tesla-elon-musk-autopilot-dmv-fsd-exaggeration

[nctnico]

If I look at the accidents that have happened with Tesla's FSD then I want to seperate these in 3 categories: ones that are unavoidable either way, ones that happened in places which are unsuitable for self driving cars and ones that could have been avoided by adapting the road. The latter case is where the low hanging fruit is and where you can get the biggest benefit in terms of increasing the capacity of roads / reducing congestion.

Remember, Tesla FSD is called Full Self Driving, but that's in name only. It's not actually full self driving. It's actual name is AutoPilot, and that's what it officially is, and advanced autopilot that can lane change and brake etc. And it has a cute auto-summon mode where it find you in a carpark or something. They are very clear about this on their website. It's Musk that has been personally talking up "Full Self Driving" and fully autonymos for years now, or at least implying it.
He's either had a technicaly awakening or the suits have tapped him on the shoulder to calm it down, they aren't even remotely close to full autonymous self self drivign whcih Musk now has essentially admitted.

https://www.theverge.com/2021/5/7/22424592/tesla-elon-musk-autopilot-dmv-fsd-exaggeration

Yep. I just called it 'Tesla's FSD' in order to use the same terminology as Tesla does. A while ago someone else has posted links to official documents basically saying that Tesla is not going to develop their 'Autopilot' much further. IIRC Tesla is actually at a lower level compared to Waymo and some others. So yes, there is a lot of hype from Musk going on.

[sandalcandal]

Remember, Tesla FSD is called Full Self Driving, but that's in name only. It's not actually full self driving. It's actual name is AutoPilot, and that's what it officially is, and advanced autopilot that can lane change and brake etc. And it has a cute auto-summon mode where it find you in a carpark or something. They are very clear about this on their website. It's Musk that has been personally talking up "Full Self Driving" and fully autonymos for years now, or at least implying it. There are countless FSD beta video out there showing how dumb the system is in even basic almost ideal condition scenarios.
He's either had a technical awakening or the suits have tapped him on the shoulder to calm it down, they aren't even remotely close to full autonymous self self driving which Musk now has essentially admitted.
https://www.theverge.com/2021/5/7/22424592/tesla-elon-musk-autopilot-dmv-fsd-exaggeration

AFAIK FSD and Autopilot are actually different products. Autopilot is driving assistance which is already common in industry and Full Self-Driving is an upgrade/add-on meant to be exactly that.
https://thenextweb.com/news/differences-between-teslas-autopilot-and-full-self-driving-system-adas-tech
https://insideevs.com/news/501634/tesla-model3-autopilot-fsd-features/

Can you share some of the dumb videos?

Edit: The "Full Self-Driving Capability" add-on for Model S is $AU10.100 and includes the MV processing hardware "FSD computer" and unlocks the Navigate on Autopilot, Auto Lane Change, Autopark, Summon and Traffic Light and Stop Sign Control functions. Cost is same on Model 3 but FSD computer is already installed. "Autosteer on city streets" is listed as upcoming and seems to be what's referred to in the "FSB beta" releases with people driving around on the streets with the car steering itself.

Apart from "Traffic Light and Stop Sign Control" most features seem common with other high end smart vehicles. Only BMW and Cadillac seems to have the full set.
https://www.bmw.com/en/innovation/the-main-driver-assistance-systems.html
https://www.volkswagen.com.au/en/technology/advanced-driver-assistance-systems.html
https://www.ford.com.au/suv/puma/safety/
https://asia.nikkei.com/Business/Automobiles/Toyota-unveils-new-cars-in-push-for-advanced-driver-assist-tech
https://www.cadillac.com/world-of-cadillac/innovation/super-cruise (added after bdunham7 pointed it out)

[sandalcandal]

So I was having a think about issues with LIDAR/RADAR based systems. One of the main "false triggers" for radar is apparently increased returns due to things in the environment that aren't actually in the way of the vehicle e.g. an overpass. You could probably solve this by using PN (pseudo noise) radar or RM (random modulation) CW LIDAR (which have been a concepts since the late 70s) to over come range ambiguity, Doppler ambiguity, and clutter. You could even use multiple antennas then process correlations to get pretty robust 3D maps. Needless to say such systems introduce much more complexity and I'm not an expert in these technologies.

I'm not really sure what people have actually implemented but it doesn't seem like these approaches are being used? Are they potentially restricted tech?

Most LIDAR looks like single point disco balls. Haven't seen anything that looks like multiple radar antennas either (can you get wavelengths small enough for multiple points?).

Impossible to tell from pictures if PN/RM is being used and they aren't mentioned in press releases I could find. It is mentioned in some fairly recent papers though
https://arxiv.org/pdf/1909.01729.pdf
https://www.researchgate.net/publication/338506276_Pseudo-LiDAR_From_Visual_Depth_Estimation_Bridging_the_Gap_in_3D_Object_Detection_for_Autonomous_Driving

I'm guessing at a certain level of complexity in hardware and data processing, it becomes more efficient (cost and development wise) to simply use multiple visual image sensors and neutral net based algorithms to perform processing?

[bdunham7]

You note that I said "in an ideal world".  In the real world an automated system drives into to the back of a truck and it makes world news.  I personally know two human drivers who have done it and it didn't even make local news (both survived, but long hospital stays and months to a not quite complete recovery)..  The total number of human drivers who have performed this feat has to be in the thousands.

I actually believe it is the lawyers who will end up deciding this, but can't predict which group of lawyers will win.  The ones sueing car companies for fielding a less than perfect autonomous system, or the lawyers sueing car companies for not fielding an autonomous system that would have prevented a stupid driver death.

I think demanding that autonomous systems performance greatly exceed humans statistically is not the wrong decision, even in an ideal world.  Sure, some of that is driven by humans thinking that even if some other idiot drives into the back of a truck, they won't.  But autonomous systems should work with humans to achieve greater safety, not replace them as a publicity stunt or gimmick.  There are driving assistance systems that do this quite effectively, although not perfectly yet.  Automatic Emergency Braking from Volvo is one example that won't drive you into a truck, but it will prevent you from doing it 99% of the time.

Expectations of automated systems are not based on human fallibility at the same task.  If I forgetfully burn my dinner on the stove and start a fire, that's human error.  If an electric teakettle doesn't shut off and the thermal cutout doesn't work and a fire ensues, the manufacturer will be sued and they'll likely lose and unless shown to be a one-off, the product will be recalled.  That failure is not acceptable even in small numbers.

[bdunham7]

Also LIDAR and radar systems are allegedly too prone to interference in real-life practical conditions. Apparently "phantom braking" which is an issue with driving safety/cruise control systems from multiple automakers https://www.baileyglasser.com/news-lawsuit-targets-nissan-and-infiniti-vehicles-for-faulty-phantom-braking is caused by radar interference or false signals. This is the reason I see cited for choosing MV as a more robust means of achieving safe self-driving.

I wouldn't cite allegations by a lawyer in a press release seeking clients if I expected to be taken seriously.  AEB systems aren't perfect either, but they don't have to be to be a great improvement in safety.  I think the best ones will synthesize different sensors.  Lidar and radar are important because they can allow the systems to see what humans can't.  If the law firms allegations are correct, that system needs some work.  However, Lidar/radar/synthesized AE B systems currently exist and actually seem to work pretty well in practice.  Volvo has had an AEB system of one sort or another for over a decade and it has been shown to improve safety.  Of course, that doesn't mean they can be extended into an acceptable autonomous full self driving system.

[bdunham7]

Start with full autonomic driving on highways & main roads in/outside cities and use the system as an anti-collission / smart cruise control aid in non-autonomous driving areas.

GM has essentially that right now in their Cadillac Super Cruise system.  And if you can wait until 2023, they'll have an electric SUV (ESUV?  EV-SUV?) with it.  Or so they say.

https://www.cadillac.com/electric/lyriq

[sandalcandal]

Start with full autonomic driving on highways & main roads in/outside cities and use the system as an anti-collission / smart cruise control aid in non-autonomous driving areas.

GM has essentially that right now in their Cadillac Super Cruise system.  And if you can wait until 2023, they'll have an electric SUV (ESUV?  EV-SUV?) with it.  Or so they say.

https://www.cadillac.com/electric/lyriq

Looks the same or slightly less functional than the currently publicly released Tesla FSD

[Jan Audio]

They cant even make the tom tom route planner correct.
Trucks go in small villages, because tom tom say so.
Smart systems, No brains.

[nctnico]

They cant even make the tom tom route planner correct.
Trucks go in small villages, because tom tom say so.

Well, TomTom is pretty bad. I once ended up driving on a dirt road through a vineyard somewhere in France. Google maps OTOH is pretty good. I have a feeling that Google is creating routes based on where local people are driving mostly.

[bdunham7]

Looks the same or slightly less functional than the currently publicly released Tesla FSD

No, it's not even remotely close in how it works nor what it tries to do.  It only works as a mostly-autonomous system on select roads which have been preapproved and mapped by GM.  It also uses LIDAR and other sensors.  The idea is that its reach doesn't exceed its grasp.

[Marco]

Also LIDAR and radar systems

LIDAR and radar are entirely different beasts. Radar zones are so f'ing huge for the moment they have to make wild guesses about stationary objects in the zone (among which the road). Collision avoidance systems are tuned towards ignoring false positives to the point they miss a lot, like stationary firetrucks in the case of Tesla. 60+ GHz 3D imaging radar similar to LIDAR is theoretically possible, but not what cars have.

So since radar is useless without human backup, to actually determine distance of objects in view what remains are LIDAR, stereo and multiframe image analysis. Tesla uses the last one, but it's by far the most fragile ... yes humans can do it quite well, but we can apply common sense and the car can only apply artificial stupidity.

Personally I think stereo is a perfectly reasonable alternative to LIDAR ... but multiframe image analysis is just stupid.

[sandalcandal]

Personally I think stereo is a perfectly reasonable alternative to LIDAR ... but multiframe image analysis is just stupid.

Stereo imaging? Isn't that just a subset of multiframe image analysis?