Driverless taxi service getting approved in SF

[SiliconWizard]

https://www.theguardian.com/technology/2022/jun/03/california-driverless-taxi-cars-san-francisco

[eti]

https://www.theguardian.com/technology/2022/jun/03/california-driverless-taxi-cars-san-francisco

No driver to pay, so costs can be slashed to 1/3rd I reckon would be fair. Let the dimwits of SF test them out! 

[Stray Electron]

https://www.theguardian.com/technology/2022/jun/03/california-driverless-taxi-cars-san-francisco

No driver to pay, so costs can be slashed to 1/3rd I reckon would be fair. Let the dimwits of SF test them out! 

   The drivers don't control the cab companies in SF and I'm sure the drivers are 100% against this for their own reasons. But I can see why the cab companies would be in favor of it.  That said, I think that in People's Republic of California it would be political suicide for a politician to vote in favor of a bill like this that would cost tens of thousands of people (the drivers) their jobs and probably cost a lot of pedestrians their lives. Google, Uber, Apple and the other companies that are pushing the adoption of this kind of technology don't own the State of California. Yet!

[SiliconWizard]

What's interesting is how the world (and it's not just California, even though a lot of those "new things" are happening there first) has sneakily forgotten about the precautionary principle, and is interpreting it in a "creative" manner:

The ride-hailing service initially will consist of just 30 electric vehicles confined to transporting passengers in less congested parts of San Francisco from 10pm to 6am. Those restrictions are designed to minimize chances of the robotic taxis causing property damage, injuries or death if something goes awry. It will also allow regulators to assess how the technology works before permitting the service to expand.

So instead of waiting for more thorough testing and analysis before granting approval, regulators now tend to deliver "temporary approvals" to let the services or products be tested in the field, in limited number, using regular people as guinea pigs. Fascinating.

[james_s]

A friend of mine predicted years ago that the driverless car thing would be pushed out hastily and then at some point there would be an event or series of events that kills a bunch of people and the whole thing is outlawed, maybe this will play out that way.

[SiliconWizard]

A friend of mine predicted years ago that the driverless car thing would be pushed out hastily and then at some point there would be an event or series of events that kills a bunch of people and the whole thing is outlawed, maybe this will play out that way.

Maybe so, but I'm more inclined to think that the actual scenario will be a few accidents that will be seen as nothing more than collateral damage, all the benefits being heavily praised, and that will become part of everyone's daily life. It's even likely that even if there are some catastrophic accidents, people will get bombarded with statistics showing that it has saved way more lives than it has killed.

[james_s]

Maybe so, but I'm more inclined to think that the actual scenario will be a few accidents that will be seen as nothing more than collateral damage, all the benefits being heavily praised, and that will become part of everyone's daily life. It's even likely that even if there are some catastrophic accidents, people will get bombarded with statistics showing that it has saved way more lives than it has killed.

But who is responsible for those accidents and how much is a life worth? A few accidents that result in death could sink a company, whereas conventional accidents will be blamed on the individual that caused the accident. Then of course there will also be vandalism and people trolling the things for fun, while the hapless passenger sits helplessly inside.

[Zeyneb]

What gives me confidence in a safe taxi service is that the drivers life is at stake as well. And if the driver owns its taxi even better because then (s)he will be careful not making minor accidents as well.

[Ed.Kloonk]

What gives me confidence in a safe taxi service is that the drivers life is at stake as well. And if the driver owns its taxi even better because then (s)he will be careful not making minor accidents as well.

Come to Sydney. Get a taxi from the airport to your hotel.

[magic]

A friend of mine predicted years ago that the driverless car thing would be pushed out hastily and then at some point there would be an event or series of events that kills a bunch of people and the whole thing is outlawed, maybe this will play out that way.

By 2030 somebody in the US will come up with a small suitcase which converts one those into a cruise missile

[eti]

If a human driver kills people, there is accountability. If SOFTWARE does so, it is waved away as "It was an unfortunate accident, <blah blah rehearsed corporate 'apology' (ass-covering)> and we will take extra steps..." blah blah blah fking blah>

They want ALL the profit, but none of the accountability, and a robot can't be sent to prison.

[Ed.Kloonk]

If a human driver kills people, there is accountability. If SOFTWARE does so, it is waved away as "It was an unfortunate accident, <blah blah rehearsed corporate 'apology' (ass-covering)> and we will take extra steps..." blah blah blah fking blah>

They want ALL the profit, but none of the accountability, and a robot can't be sent to prison.

True, but driverless cars are an extension of automated machinery. Was it safe? How come it wasn't safe? Who made it unsafe? 

....

[Stray Electron]

If a human driver kills people, there is accountability. If SOFTWARE does so, it is waved away as "It was an unfortunate accident, <blah blah rehearsed corporate 'apology' (ass-covering)> and we will take extra steps..." blah blah blah fking blah>

They want ALL the profit, but none of the accountability, and a robot can't be sent to prison.

  This pretty much sums up the whole issue.  Ask yourself, has anyone gone to jail yet over the two Boeing 737Max crashes?  And how many people died in those accidents?

[jonpaul]

Besides the crime, filth, homeless,  risques of Getting hit by a driverless car or taxi is just  another reason to avoid the city altogether.

jon

[pcprogrammer]

A friend of mine predicted years ago that the driverless car thing would be pushed out hastily and then at some point there would be an event or series of events that kills a bunch of people and the whole thing is outlawed, maybe this will play out that way.

What is hastily? Back in 1995-1998 I was involved in developing a driver-less container transport system called COMBI-ROAD. https://www.sae.org/publications/technical-papers/content/981940/ At some point in time there was a demonstration of this system scheduled to happen at the same time an American company (don't remember the name) came over to do a demo of a driver-less car. There was this new highway, just finished and not in use yet, reserved for this. (Think it was a section of the A50) The cars drove themselves in series but with a "non driver" behind the wheel to intervene in case of a problem. The faith in the system was not very high

The demonstration for the COMBI-ROAD system was conducted at the same time on a company terrain (At the time I think named Traxis) For the time that the truck was on it's own track there was no driver. It did work quite well, but to be honest the obstacle detection had it's weak points. A person on the track was only detected when the infrared beams were interrupted, so there were spots that the truck would happily smush anyone on the track

So now some 25 years later the self driving car is still not safely there

Sure with AI and cameras a lot is possible, but failure is always there. Humans are unpredictable and with emotions in play there are so many scenarios that can go wrong. Let's be honest we, as drivers in general, are not capable of driving without accidents. Only in a perfect setup where there are no other users on the road self driving cars might do well.

The point of responsibility is very true. Whom to blame if it goes sour

[james_s]

A friend of mine predicted years ago that the driverless car thing would be pushed out hastily and then at some point there would be an event or series of events that kills a bunch of people and the whole thing is outlawed, maybe this will play out that way.

What is hastily? Back in 1995-1998 I was involved in developing a driver-less container transport system called COMBI-ROAD. https://www.sae.org/publications/technical-papers/content/981940/ At some point in time there was a demonstration of this system scheduled to happen at the same time an American company (don't remember the name) came over to do a demo of a driver-less car. There was this new highway, just finished and not in use yet, reserved for this. (Think it was a section of the A50) The cars drove themselves in series but with a "non driver" behind the wheel to intervene in case of a problem. The faith in the system was not very high

The demonstration for the COMBI-ROAD system was conducted at the same time on a company terrain (At the time I think named Traxis) For the time that the truck was on it's own track there was no driver. It did work quite well, but to be honest the obstacle detection had it's weak points. A person on the track was only detected when the infrared beams were interrupted, so there were spots that the truck would happily smush anyone on the track

So now some 25 years later the self driving car is still not safely there

Sure with AI and cameras a lot is possible, but failure is always there. Humans are unpredictable and with emotions in play there are so many scenarios that can go wrong. Let's be honest we, as drivers in general, are not capable of driving without accidents. Only in a perfect setup where there are no other users on the road self driving cars might do well.

The point of responsibility is very true. Whom to blame if it goes sour

Driving on a closed highway is an entirely different situation than coexisting with human drivers on public roads. This attitude is exactly the sort of thing that leads to hasty deployments, people are overly confident, the did it on a closed circuit so they think it will be easy to scale up to deal with real world driving on public roads. It's not.

[jonpaul]

Bonjour, a few questions please!

What's an acceptable system error rate..1%?  0.01%

 how much dégradation will occur due to:

1. poor or missing internet connection or poor mobile cell coverage

2. stale maps eg due to recent roadworks?

3. undisclosed or unknown Software and firmware bugs?

4. Global position error...faux signal, eg intentionally due to conflict or war?

Already just within a few years of limited testing, some deaths and serious accidents

Who exactly is responsible?

who has rights and access to the stored data from the vehicles involved?

Vehicle manufacturer, driver, owner?


Bon courage


Jon

[pcprogrammer]

Bonjour a vous,

All very good questions which are hard to answer.

What is safe error response? Just stopping on the middle of the road if for instance the internet connection drops is not safe and certainly a hinder to the rest of the users on the road.

About GPS, it does not work when in a tunnel or underpass, so can't rely on it. Radar does not look around corners and can be absorbed so again nothing to rely on. Etc, etc.

Way to many problem points with this for me to get into a self driving car

Cordialement,
Peter

[jonpaul]

Rebonjour Cher Peter,

any détection system, radar lidar, sonar has limited capture angle, and update rate.

false negatives can be as bad in concequence as false positives...

The complexity of the many computers and 100s mega lines of SW make a perfect detection impossible.

I would be surprised if the serious incident error rate were better than 0.1% = one failure per 1000 incident.


"No 9000 computer has ever made a mistake or distorted information....."

HAL9000 in Stanley Kubrick's classic 1968 film, "2001, A Space Odyssey"

bon courage,

jon

[pcprogrammer]

"No 9000 computer has ever made a mistake or distorted information....."

Bonjour Jon,

that was before Windows came along

About the detection systems, that is why the COMBI-ROAD system I talked about was using infrared beams to transfer information about the road including wetness and temperature and obstacles for the next kilometer ahead to the vehicle. The safety in the system was the vehicle stopping when there was a problem, and all vehicles behind it would too. Can only work on a dedicated and separate road. But it also has limitations, like railroads also have and thus need a way to keep people of the track.

The wife is calling supper so till next post,

Greetings,
Peter

[rstofer]

If a human driver kills people, there is accountability. If SOFTWARE does so, it is waved away as "It was an unfortunate accident, <blah blah rehearsed corporate 'apology' (ass-covering)> and we will take extra steps..." blah blah blah fking blah>

They want ALL the profit, but none of the accountability, and a robot can't be sent to prison.

  This pretty much sums up the whole issue.  Ask yourself, has anyone gone to jail yet over the two Boeing 737Max crashes?  And how many people died in those accidents?

You can only go to jail if you are convicted of a CRIME and that can only occur if there was intent or gross negligence.  An accident is just that; an accident.  In most cases, there are so many contributing factors that it is impossible to determine negligence and proving intent is even harder.

Now, that doesn't mean there won't be civil lawsuits, that's a given.  But nobody goes to jail over a civil judgement.

There is still plenty of work to do on the AI driving the car.  Right now it has a propensity for driving into police cars and getting confused with traffic cones.

https://www.autoweek.com/news/green-cars/a37425353/another-tesla-hits-police-car/




When you ride in a self-driving car, your life is not only in the hands of a programmer but also a math major.  The day won't come when I let a car drive me around.  Not happening!

It is estimated that it takes 70,000 GPU hours to rebuild the Tesla neural network.  I'm still trying to find a definition for GPU-hour since GPU apparently refers to the entire device rather than the highly variable number of CUDA cores in the device.  I have one graphics card with a couple of hundred CUDA cores and another with nearly 6000.  Both are GPUs (of a sort) but they certainly aren't equivalent.

[SiliconWizard]

We had a number of threads about AI and liability, those points are definitely valid and we could already see this is a very polarizing topic.
Yes, AI in everything is a great way of abolishing the concept of liability.

But what got me even more concerned in this particular case is how authorities are dealing with this.
As I said above, it would appear that the "new normal" is to put the precautionary principle aside and approve things that aren't formally proven safe, just because there is now a sense of emergency that would justify it and that we are all supposed to embrace.

[jonpaul]

consider the situation vs airline flights crashes investigation.

Radar, grounds radio altimeter, ILS, control towers, extensive weather and runway conditions..

Still every year some failures and crashes.

Predicted much worse for autonomie véhicule
s.

Jon

[Bud]

We had a number of threads about AI and liability, those points are definitely valid and we could already see this is a very polarizing topic.
Yes, AI in everything is a great way of abolishing the concept of liability.

But what got me even more concerned in this particular case is how authorities are dealing with this.
As I said above, it would appear that the "new normal" is to put the precautionary principle aside and approve things that aren't formally proven safe, just because there is now a sense of emergency that would justify it and that we are all supposed to embrace.

What kind of emergency do you refer to in regards to taxi service, or to generic autonomic passenger vehicle for that matter? 

[Bassman59]

"No 9000 computer has ever made a mistake or distorted information....."

HAL9000 in Stanley Kubrick's classic 1968 film, "2001, A Space Odyssey"

That statement is correct in the universe of the film.

HAL did not make a mistake when it removed the life support from the sleeping astronauts. HAL did not make a mistake when it killed Frank and when it tried to kill David.

HAL's primary goal was to ensure the success of the mission to Jupiter. As part of its continuing operation and calculation, it made a determination that the fallibility of the humans put that mission in jeopardy. Thus the only logic action was to eliminate the human element.

You might remember that HAL had a twin on Earth, and a test was run, comparing HAL to that twin, and there was no evidence of malfunction. The test they didn't run was to ask the Earth-bound twin to "be in HAL's shoes on the Discovery," and get its response. Would the twin have made the same fatal decision? Obviously yes.

Remember that Clarke was clearly aware of Asimov's Three Laws of Robotics. But HAL was not a robot -- specifically, HAL was not sentient. Though maybe Clarke's 4th Law should have been "Any sufficiently-advanced processing power/programming is indistinguishable from sentience" ?