Ford commits to fully autonomous vehicles within five years

[akirby]

They can be successful in specific, controlled environments such as parking lots, stop and go traffic, etc. but not for everyday driving everywhere.

[jasonj80]

Talked to a Technologist/Ethics that works at one supplier developing these systems and were having a discussion over this topic. Here is an example that he gave that they still don't have an answer to and may never to the point it could actually may end up being up to the driver to select options configuring the system. A child gets out of a parked car and immediately bolts across a busy street the computer calculates that if the child is hit they stand a 95% chance of death, but the occupants of the vehicle will be unharmed. The best option the computer calculated is to swerve and crash into an object where there is a 30% chance the occupants of the car are killed and 95% they will be injured but the child would be unharmed. Which does the computer decide? You as the driver may select save me at all costs, or use a dial and increase the chance of death or injury if there is an unavoidable accident. Now going to court they will be documented data that you put yourself in front of others and how will that play out?

Also how do city's makeup for the revenue shortfall from no tickets that a driverless car would have? When you have even small cities budgeting $1,000,000 in ticket revenue where is that made up?

[akirby]

It's funny that the people who are the most skeptical about this are the ones with extensive computer programming experience.....

Edited August 22, 2016 by akirby

[fuzzymoomoo]

It's funny that the people who are the most skeptical about this are the ones with extensive computer programming experience.....

I'm just as skeptical as you and I'm in the generation that's "supposed" to put all of my faith in technology

[fordmantpw]

It's funny that the people who are the most skeptical about this are the ones with extensive computer programming experience.....

 

Where have I seen that before....

[mackinaw]

When President John F. Kennedy made this famous proclamation in 1961, "I believe that this nation should commit itself to achieving the goal, before the decade was out, of landing a man on the moon and returning him safely to earth," many aerospace engineers doubted this was possible. Eight years later, man walked on the moon.

 

Autonomous cars will happen. The question is when.

[RichardJensen]

When President John F. Kennedy made this famous proclamation in 1961, "I believe that this nation should commit itself to achieving the goal, before the decade was out, of landing a man on the moon and returning him safely to earth," many aerospace engineers doubted this was possible. Eight years later, man walked on the moon.

 

Autonomous cars will happen. The question is when.

 

With all due respect, that was a much easier task. There were far fewer variables, and there was no great leap in computational ability required to achieve that goal.

 

Consider that sending a man to the moon was largely a question of:

 

- whether sufficient boost could be generated

- whether components would be sufficiently durable and reliable

- whether necessary systems could be controlled effectively by a crew of three and for part of the trip, two.

 

People need to stop pretending that the moon landing was, due to its spectacular nature, an undertaking that required significant technological breakthroughs. It manifestly did not. It required scaling up existing technologies: larger rockets, more durable flight suits & cockpits, more durable electronics. And even at that, the undertaking cost billions.

 

To make it comparable, an autonomous car would require not a breakthrough in artificial intelligence, but just better and cheaper sensors, etc.

Edited August 22, 2016 by RichardJensen

[BORG]

Self driving will likely be a very slow evolutionary process to the point it will resemble rail cars more than anything else (confined to small areas like Ford's system). The rest of it will likely act as safety or 'advanced cruise' driving aids. I think GM's "Super Cruise" is actually the best way to describe what automation will ultimately look like for the rest of us. So I think this is where the focus should be, not in marketing stunts which ultimately confuse and endanger people.

 

Software certainly has the power to make more consistent and better researched decisions than a significant number of drivers who may or may not have a developed sense of the road and the possible outcomes.

 

This video is actually a perfect all-encompassing example of why automation is key to safety.

 

A computer would not make this sort of decision for example because the maneuver is done without enough visual data to make assumptions without significant risk. And more importantly, the risk was likely taken because of a navigational error by the driver without an advanced understanding of road conditions (location of the exist or traffic) which would not happen in automation. Furthermore, since risk avoidance is a known human inconsistency, it's best handled by automation in which risk avoidance is paramount. And should automation fail in any specific instance, the entire system adapts but humanity has no such mechanism for correction. Therefore, the only way to improve safety is to change the technology, not the human...which has always been the case.

 

Edited August 22, 2016 by BORG

[RichardJensen]

A computer would not make this sort of decision for example because the maneuver is done without enough visual data]

 

Despite anecdotal evidence such as the above, as I pointed out earlier, crashes alone (not injury accidents or fatal accidents) are so rare as to be irrelevant statistically.

 

Providing anecdotal evidence of poor driving does not further the discussion, as it is not balanced by proportionally overwhelming accounts of accident avoidance.

 

For this video to fit into context, you would have to demonstrate that an autonomous vehicle would also be able to avoid accidents that average drivers routinely avoid, as well as this accident (which, despite your assurance, could happen with an autonomous vehicle that was relying on obsolete mapping data).

Edited August 23, 2016 by RichardJensen

[fordmantpw]

Should I provide anecdotal evidence of a software bug to bolster the case for non-fully autonomous vehicles? Maybe a few edge cases like the video above?

[mackinaw]

 

With all due respect, that was a much easier task. There were far fewer variables, and there was no great leap in computational ability required to achieve that goal.

So landing a man on the moon was easy? Really? I’d like to be there when you tell Buzz Aldrin that.

You’re missing the point. Many in NASA said “no way” when JFK made that famous proclamation so many years ago. But the best and the brightest got together and made it happen.

This is exactly the case with autonomous cars. They will happen. Only a luddite would think otherwise. Engineering problems that appear insurmountable now, will be solved. Do you realize that it’s only been 40 years since Steve Wozniak designed the Apple 1 computer? Look at the advances in technology since.

I have no interest in owning an autonomous car, I enjoy driving too much, but in 20 years when I’m truly old, the idea of being able to hop into an autonomous car and drive anywhere I want, holds great appeal.

This forum is so conservative. So afraid to dream. Always playing it safe. I really don’t know why I come here.

 

[4d4evr-1]

man was not meant to fly, if he was God would have given him wings.....

[4d4evr-1]

I believe it will begin just as Uber and Ford have in mind. In a 'controlled' area there will be a site where you can request a vehicle to arrive at a certain time for your appointment. The service will respond with confirmation and pick you up at designated time and place and will deliver you to your destination at desired time and you get out. The car then proceeds to its next appointment for next delivery. Since every car can be used virtually 24/7 except for refueling and maintenance many fewer cars will be needed and virtually no parking used.

This is how it begins and works like taxis with no drivers needed. Testing will be at slow city speeds and with drivers in other cars, There will be problems just like with every human interaction in life.

To think we are just staying where we are is not very intelligent to me.

Can I make this work, no, but I have issues just typing this post.

[BORG]

 

Despite anecdotal evidence such as the above, as I pointed out earlier, crashes alone (not injury accidents or fatal accidents) are so rare as to be irrelevant statistically.

 

Providing anecdotal evidence of poor driving does not further the discussion, as it is not balanced by proportionally overwhelming accounts of accident avoidance.

 

For this video to fit into context, you would have to demonstrate that an autonomous vehicle would also be able to avoid accidents that average drivers routinely avoid, as well as this accident (which, despite your assurance, could happen with an autonomous vehicle that was relying on obsolete mapping data).

 

 

This discussion is not based upon your assumption or belief that accidents are statically irrelevant and that autonomous technology is a static form of safety technology that does not change over time. There are 30,000+ fatal car accidents in the US a year, that video is an example of how that happens (although she actually survived).

 

As for changing map data, you do understand however that autonomous driving is a multivariable system and map data is only one part of the system? Or the fact that map data is regularly updated in a connected system (and that significant infrastructure changes don't go unnoticed).

 

There is also the simple fact that as a consumer technology, people LIKE this technology and there is nothing wrong with the convenience of it assuming it works safely.

Edited August 23, 2016 by BORG

[fordmantpw]

 

So landing a man on the moon was easy? Really? I’d like to be there when you tell Buzz Aldrin that.

 

Compared to fully-autonomous cars? Yes, yes it was. So, that was 9 years from saying it will happen to landing on the moon? How long have people been working on autonomous cars...with more computing power?

 

 

You’re missing the point. Many in NASA said “no way” when JFK made that famous proclamation so many years ago. But the best and the brightest got together and made it happen.

 

So, the point is that just because somebody says something, it will happen? OK, I predict I will wake up with 1 million dollars in a pile on my living room floor in the morning! Let's get together and make it happen!

 

 

Engineering problems that appear insurmountable now, will be solved.

 

It's not an engineering problem...it's a 'fact of life' problems. Machines are not capable of thinking and reasoning the way humans do. It's just not possible. Progress has been made, yes, but it's still a defined and given set of circumstances. You have given parameters that have to be met. The real world isn't that way. In a very controlled situation, yes, fully-autonomous is possible. But to be able to just get in a car and go ANYWHERE without needing to do anything? Nope. Not in my lifetime...and I'm 40 so I hope to live a good long while yet. It's not being conservative, it's just being a realist about the way the world works. Hell, how would that car handle the new water breaks I just put in the hillside of my driveway to keep the water from washing?

 

If you want an example, try a piece of speech to text software. You know how it has to practice and learn your speech, and how it's different from everyone else's? That's pretty simple compared to driving. But with driving, you don't get 'practice'. If you fail on the first attempt, you kill people.

[fordmantpw]

 

 

This discussion is not based upon your assumption or belief that accidents are statically irrelevant and that autonomous technology is a static form of safety technology that does not change over time. There are 30,000+ fatal car accidents in the US a year, that video is an example of how that happens (although she actually survived).

 

As for changing map data, you do understand however that autonomous driving is a multivariable system and map data is only one part of the system? Or the fact that map data is regularly updated in a connected system (and that significant infrastructure changes don't go unnoticed).

 

There is also the simple fact that as a consumer technology, people LIKE this technology and there is nothing wrong with the convenience of it assuming it works safely.

 

BORG, as much as you complained about MFT and how buggy it was, for an infotainment system, you would put your trust in the same company to drive your car down the road at 65 MPH while you took a nap? Really?

[jasonj80]

 

So landing a man on the moon was easy? Really? I’d like to be there when you tell Buzz Aldrin that.

You’re missing the point. Many in NASA said “no way” when JFK made that famous proclamation so many years ago. But the best and the brightest got together and made it happen.

This is exactly the case with autonomous cars. They will happen. Only a luddite would think otherwise. Engineering problems that appear insurmountable now, will be solved. Do you realize that it’s only been 40 years since Steve Wozniak designed the Apple 1 computer? Look at the advances in technology since.

I have no interest in owning an autonomous car, I enjoy driving too much, but in 20 years when I’m truly old, the idea of being able to hop into an autonomous car and drive anywhere I want, holds great appeal.

This forum is so conservative. So afraid to dream. Always playing it safe. I really don’t know why I come here.

 

 

Some of this isn't a Engineering decision, it is an ethical decision which is completely different, who does the car save? You have other people determining your ability to live. Do they look and say at 50 they are older we should take the chance and kill him to save the life of the youth or if you were 30 and had a family that the kid is better off to be killed as the 30 year old has 2 small children and needs parents, what if the 30 year old is single. What if one person is a Doctor, Teacher or a basic labor worker are their lives calculated different values by the computer? Who in the end writes the algorithm to determine this, as you are basically writing if XXXXXXX event happens it is a death sentence as the one person is more important than the other. Engineering and coding can be done, and will probably be getting there in 5 years, but it is the ethics in how those if/or/not are programmed that can have an impact that is so far beyond simply sending someone to the moon.

Edited August 23, 2016 by jasonj80

[RichardJensen]

 

So landing a man on the moon was easy?

 

 

 

The challenges were different; they were such that they could be readily framed. They involved mathematics, materials engineering, miniaturization and controls. They did not involve artificial intelligence.

[RichardJensen]

 

 

This discussion is not based upon your assumption or belief that accidents are statically irrelevant

 

That is neither an assumption nor a belief. That is a fact.

 

The failure rate of human piloted cars (crashes--NOT fatal or injury accidents) is less than the six sigma standard for defects in industrial processes.

 

Six sigma failure rates are ~3.4 occurrences per million; crashes occur 1.85 times per million vehicle miles driven.

 

Yes, the loss of life is tragic, but that does not mean that fatal accidents, injury accidents or even accidents in general are so widespread as to necessitate a rethink of the current vehicle control system (i.e. human driver), much less suggesting that autonomous cars would be safer.

Edited August 23, 2016 by RichardJensen

[fordmantpw]

 

That is neither an assumption nor a belief. That is a fact.

 

The failure rate of human piloted cars (crashes--NOT fatal or injury accidents) is less than the six sigma standard for defects in industrial processes.

 

Six sigma failure rates are ~3.4 occurrences per million; crashes occur 1.85 times per million vehicle miles driven.

 

Yes, the loss of life is tragic, but that does not mean that fatal accidents, injury accidents or even accidents in general are so widespread as to necessitate a rethink of the current vehicle control system (i.e. human driver), much less suggesting that autonomous cars would be safer.

 

There are decidedly more than 3.4 bugs for every 1.85 millions lines of code written...

[RichardJensen]

 

There are decidedly more than 3.4 bugs for every 1.85 millions lines of code written...

 

BTW: Why aren't we talking about fully autonomous air traffic control? There are far fewer airplanes in the air at any given point in time than there are cars on the road. There are also far fewer airports than their are miles of highway to map and manage.

 

Surely, if cars can drive themselves, then there is absolutely no technological necessity to have human beings directing air traffic in the US.

[BrewfanGRB]

Well, let's see:

 

You have a horrifically bureaucratic FAA that, like virtually every federal agency, simply cannot build and deploy significant computer systems. (And I say this as someone that LIKES the federal government).

Then you have the FAA's charter that forces them to satisfy opposing forces (safety and development/promotion of commercial air travel--well, all air travel, I suppose)

Then you have airlines that complain about the current system that is clearly untenable on a going forward basis but want to pay little or nothing to improve their own aircraft or toward or for the new system infrastructure.

Then you have the need to update and improve airports--some that are as nearly incompetent (or more) than the FAA--who also don't want to pay

A despicable Republican-controlled Congress that doesn't want to spend any money--not because they have principles of fiscal conservatism (they had no problem at all prosecuting 2 wars entirely using debt while reducing the tax burden on the rich), but because doing so either gets them primaried or confers a "win" on the President. (To be fair, this is a smaller factor given the fact the FAA can't get anything done when they are given the money).

 

But DESPITE all this, there are improved system already in place at LAX and I believe ATL, among others that significantly increase the number of operations per hour.

 

But this is STILL a terrible analogy because planes do not NEED to conflict like traffic does. You really remove the ethical dilemmas of autonomous cars here plus you remove the difficult parts of navigation (how do you drive on a dirt road? How do you park the car closer to the front porch so grandma can get out?)

 

Keep in mind, 95% of each flight is already automated. The routing is anachronistic and once the FAA and the airlines get off their ass and go fully GPS, you'll just have more direct routings that are more efficient and shorter. They'll still have spacing, but realistically, given the size of the airspace, that's not a big deal. It matters on approach and departure. If you can space the aircraft closer, you'll be able to improve airport efficiency.

 

At airports with full ILS, aircraft can land themselves already anyway. It's not done because on approach and departure is when the aircraft is at greatest risk.

 

Given we've established human drivers are already better than six sigma levels of safety, but flight is orders of magnitude safer than driving and we've now decided that is the standard to which flight will be held, pilots will be needed for a very long time.

 

But here's the deal: Yes, there's less traffic. Yes, there's more space. But if you're driving and something goes wrong, you can just stop and pull over. There are still only a relatively small number of things that can go wrong. In flight, there are many, many more things that can go wrong, but most importantly, you can't just stop (or even just decide to land randomly).

 

If we think we can't get AI to work through everything on the ground and make ethical decisions, etc then I think flight is exponentially harder. Yes, pilots still do cause a lot the crashes that occur--even in the face of automation that if left alone, would've been fine (see: AF 447--OMG I WANT TO STAB SOMEONE IN THE EYE OVER THAT ONE). But pilots also save aircraft from themselves and as noted above, have problem solving skills that will be very, very hard to program. (If it's even possible).

Edited August 23, 2016 by BrewfanGRB

[fordmantpw]

But this is STILL a terrible analogy because planes do not NEED to conflict like traffic does. You really remove the ethical dilemmas of autonomous cars here plus you remove the difficult parts of navigation (how do you drive on a dirt road? How do you park the car closer to the front porch so grandma can get out?)

 

No, it's a perfect analogy, and you mentioned several reasons why flight is similar to driving:

 

 

At airports with full ILS, aircraft can land themselves already anyway. It's not done because on approach and departure is when the aircraft is at greatest risk.

 

Given we've established human drivers are already better than six sigma levels of safety, but flight is orders of magnitude safer than driving and we've now decided that is the standard to which flight will be held, pilots will be needed for a very long time.

 

But pilots also save aircraft from themselves and as noted above, have problem solving skills that will be very, very hard to program. (If it's even possible).

 

I've bolded the parts that apply directly to driving an automobile the same as flying, and why they are similar and why fully autonomous cars are not viable.

[rmc523]

And looking at aircraft....even if 95% of the flight is automated, there are still controls that a pilot could take over if need be, not a steering wheel/pedal-less vehicle that you're trusting with no option to take over.

[BrewfanGRB]

 

No, it's a perfect analogy, and you mentioned several reasons why flight is similar to driving

 

 

 

I've bolded the parts that apply directly to driving an automobile the same as flying, and why they are similar and why fully autonomous cars are not viable

 

I don't even know where to begin here. I guess I'll just say "you're right" and move on. (Note that I never said anything about fully autonomous cars being viable and a must-happen. But whatever.