Forum adverts like this one are shown to any user who is not logged in. Join us by filling out a tiny 3 field form and you will get your own, free, dakka user account which gives a good range of benefits to you: No adverts like this in the forums anymore. Times and dates in your local timezone. Full tracking of what you have read so you can skip to your first unread post, easily see what has changed since you last logged in, and easily see what is new at a glance. Email notifications for threads you want to watch closely. Being a part of the oldest wargaming community on the net. If you are already a member then feel free to login now.

Or we can draft them at 18 to kill people, yet they can't drink beer till 21.

It is all very odd.

 Easy E wrote:
Something to consider about drivers and how seriously we take the dangers of human drivers.

We let 16 year olds operate cars anywhere they want once licensed with few restrictions. However, in a commercial kitchen a 16 year old is not even allowed to use a knife with no option to be licensed for safety reasons.

Let's all think about that for a moment.

Clearly, as a society we have let convenience of travel and movement trump safety concerns about human drivers. Why are we being stricter on robots driving?

At 16 I was not allowed to be hired at an auto parts store, because you had to be 18 to use the pipe cutter (for some reason). Another parts shop couldn't hire me because they used ladders that were too tall for a 16 y/o to climb on.

And I know it's been pointed out numerous times in this thread but, despite growing numbers of laws against it, I STILL see at least half a dozen times per day: people talking on phone, with phone to ear while driving, doing makeup while driving, eating while driving (hey, I'm guilty of this one), texting while driving, and all kinds of distractions going on behind the wheel. . . The reason why the law books keep getting bigger and thicker on this subject is simple: these things cause accidents, and many times are deadly. These are all things an automated driving system would not be doing. But these automated systems are not to a point today, where they are feasible. However, I do think that when it works, it will work better than any of the above drivers.

 sebster wrote:


Your story is obviously ridiculous, and it's clear you're not even trying to think about this sensibly.

You suggest a critical software flaw that produces a significant chance of a fatality every time the car travels a short distance. ,

Yeah, about that:



Remember the Pinto? The car with a design flaw that made it burst into flames in even a low speed collision?

 BaronIveagh wrote:

 sebster wrote:


Your story is obviously ridiculous, and it's clear you're not even trying to think about this sensibly.

You suggest a critical software flaw that produces a significant chance of a fatality every time the car travels a short distance. ,

Yeah, about that:



Remember the Pinto? The car with a design flaw that made it burst into flames in even a low speed collision?

Goodness I remember reading about this car before, and thought to myself how scary it must have been to actually drive one.

 KTG17 wrote:


No its not short sighted. I am a software developer by career. I know a thing or two about developing. I also know a thing or two about software issues, and the time and money it takes to resolve them. And how management will ignore many of them due to time and cost. There will be millions and millions of lines of code written to handle god knows the amount of real life variables that can occur just driving to work. I would rather put the responsibility of getting there in the hands of a driver than a bunch of procedures written by god knows who.

What kind of software do you do? Cars will all be using something that's adhering to MISRA and an appropriate ASIL (Automotive Safety Integrity Level) level, with lots of checks, they won't just be code thrown together for a website.

My iPhone freezes, my laptop needs a reboot at times, gak happens.

They are both running an OS that needs to handle all sorts of software running on all sorts of hardware, which introduces bloat and incompatibilities.

Your car will be running an embedded OS with just the code it needs to run on that set of hardware. There will probably be a head unit that resembled a phone OS but it won't be in any way responsible for making the decisions on what the car does. It'll likely make requests about configuration changes but all the real work will be done by a 'secure' system.

And it's not a case of whether you trust a computer to drive the car, but do you trust computers to drive cars better than a random selection of people?

Actually reading about that case on wikipedia, the car wasn't really much more dangerous than any other car in its class.

It just had the unfortunate luck of being the one which got caught in the media spotlight.

Basically all the subcompact cars of the time were extremely vulnerable.

Herzlos wrote:

Your car will be running an embedded OS with just the code it needs to run on that set of hardware. There will probably be a head unit that resembled a phone OS but it won't be in any way responsible for making the decisions on what the car does. It'll likely make requests about configuration changes but all the real work will be done by a 'secure' system.

And it's not a case of whether you trust a computer to drive the car, but do you trust computers to drive cars better than a random selection of people?

A little to this point, when I bought my most recent car, I was told by the sales guy that my vehicle has 220 different computers on board. I didn't believe him. . . Finance lady said the same gak, and it wasn't until the mechanics echoed the same thing that I started to believe them.

Many of the "computers" on board modern automobiles are more akin to calculators than laptops, as in, they have one specific job and everything about them supports doing only that one thing.

 BaronIveagh wrote:
Yeah, about that:

(image snipped)

Remember the Pinto? The car with a design flaw that made it burst into flames in even a low speed collision?

Exactly. The Pinto is famous as one of the great disasters of car manufacturing. Its design flaw caused 27 deaths over a 10 year period, with 3 million cars sold. That's the level of danger that causes people to be outraged, and a car to be removed from the market with immense reputational harm to its manufacturer. In contrast, Gray Templar is arguing that self-driving cars would be expected to have such design flaws that we couldn't risk them driving to the local dealership for updates, and that such things would be routine, and that both consumers and manufacturers would accept this as the normal course of events.

Thanks for adding further details making clear how ridiculous Gray Templar's position was. I'm sure that was your intention.

 sebster wrote:
and that both consumers and manufacturers would accept this as the normal course of events.

No, but thanks for playing.

The point was (since you clearly can't figure it out so I must walk you through it like a small child) that the sort of flaw that you're claiming could never happen has happened before. Things like the Pinto have slipped through quality control and regulation. So, saying that what Grey Templar is arguing is impossible is to not only ignore the past, but also how auto manufacturers go about deciding if a recall is warranted.

"You must be this disastrous to ride'. ______

Interestingly, if the owners are likely to all die, a recall is actually less likely than if only some of them are likely to die. This is because the math for a recall goes something like this: Does the cost of a recall meet or exceed expected lawsuit costs?

 BaronIveagh wrote:

 sebster wrote:
and that both consumers and manufacturers would accept this as the normal course of events.

No, but thanks for playing.

The point was (since you clearly can't figure it out so I must walk you through it like a small child) that the sort of flaw that you're claiming could never happen has happened before. Things like the Pinto have slipped through quality control and regulation. So, saying that what Grey Templar is arguing is impossible is to not only ignore the past, but also how auto manufacturers go about deciding if a recall is warranted.

"You must be this disastrous to ride'. ______

Interestingly, if the owners are likely to all die, a recall is actually less likely than if only some of them are likely to die. This is because the math for a recall goes something like this: Does the cost of a recall meet or exceed expected lawsuit costs?

But the Pinto didn't "slip through". It passed the regulations of the time.

It is just that the regulations weren't stringent enough. None of the cars in the same class as the Pinto made at that time would come close to passing the current regulations and laws.

 A Town Called Malus wrote:

It is just that the regulations weren't stringent enough. None of the cars in the same class as the Pinto made at that time would come close to passing the current regulations and laws.

I notice you don't bring up that it did slip past Quality Control, however.

That said, do you honestly thing that regulations for self driving cars will be nearly what they need to be when they hit the street compared to what they might be 40 years after the technology matures?

 BaronIveagh wrote:
That said, do you honestly thing that regulations for self driving cars will be nearly what they need to be when they hit the street compared to what they might be 40 years after the technology matures?

Again, you are making the wrong comparison. It doesn't matter how automated cars now compare to automated cars in the future. Of course the technology is going to improve, nobody is disagreeing with that idea. But the question that matters is how the current technology compares to the known poor safety of human drivers. Automated cars can have safety problems and kill people but still be an improvement over human drivers, and if that improvement exists they should become mandatory even if they aren't perfect yet.

 Peregrine wrote:
But the question that matters is how the current technology compares to the known poor safety of human drivers.

Not good. They're already prohibited in my workplace, despite the swarms of driven vehicles in the facility.

 BaronIveagh wrote:
The point was (since you clearly can't figure it out so I must walk you through it like a small child) that the sort of flaw that you're claiming could never happen has happened before.

But what Gray Templar is claiming hasn't happened. You used the Pinto because it is famous for being a death trap, but a look at its record shows how little increased danger a car has to have before we see a massive consumer and legal reaction, which helps us establish how ludicrous the scenario Gray Templar lays out actually is.

The Pinto's design flaw caused 27 deaths over ten years, from 3 million units sold. So if you bought one and drove it around for 10 years, you had a 27 / 3,000,000 chance of suffering from that particular flaw. That increase in death was a big deal that trashed the manufacturer's brand and forced recalls and lawsuits.

Gray Templar is trying to argue that future automated cars will suffer from such incredible design flaws that when identified when won't even dare to put them on the road to return them to the dealer. He is talking about a level of failure several orders of magnitude greater than the Pinto. And then his argument starts to get crazy, because he is arguing that in the face of such chronic product failure we won't see recalls and lawsuits. Instead he is arguing consumers will shrug, and accept that cars which are routinely found to be too dangerous to take themselves to the dealership will just have to be updated without that trip, instead being updated by remote access. He thinks we'll find out our cars are so deadly they can't be driven for update, but we'll just shrug and continue to let them drive around once updated. And he thinks these kinds of updates will be routine.

Seriously, Gray Templar's argument is not the argument you want to try and fight for. Give this one up.

Automatically Appended Next Post:

 BaronIveagh wrote:
That said, do you honestly thing that regulations for self driving cars will be nearly what they need to be when they hit the street compared to what they might be 40 years after the technology matures?

You missed the point completely. The argument isn't whether self driving cars will improve. Of course updates and improvements will happen. The point is that the original version will still be safe enough to drive that it drive to the dealership to receive updates.

 Peregrine wrote:

 BaronIveagh wrote:
That said, do you honestly thing that regulations for self driving cars will be nearly what they need to be when they hit the street compared to what they might be 40 years after the technology matures?

Again, you are making the wrong comparison. It doesn't matter how automated cars now compare to automated cars in the future. Of course the technology is going to improve, nobody is disagreeing with that idea. But the question that matters is how the current technology compares to the known poor safety of human drivers. Automated cars can have safety problems and kill people but still be an improvement over human drivers, and if that improvement exists they should become mandatory even if they aren't perfect yet.

You forgot to sling a few digs about cell phone use and the like. Here's the beef, and one you ignore. Not everyone texts while driving. I sincerely doubt that it hits 50% amongst drivers over the age of 21. The flaw that allowed the AI car to kill that woman is in EVERY SINGLE AUTOMATED CAR USING THAT PROGRAM. Do you see the issue now? Humans do stupid things from time to time, that much is a given even before pointing out Tide Pods, but you don't have all humans doing the same stupid thing that results in fatalities. With AI, if that problem exists, it will exist wholesale. THAT is the issue that you refuse to accept.

 Just Tony wrote:
You forgot to sling a few digs about cell phone use and the like. Here's the beef, and one you ignore. Not everyone texts while driving. I sincerely doubt that it hits 50% amongst drivers over the age of 21. The flaw that allowed the AI car to kill that woman is in EVERY SINGLE AUTOMATED CAR USING THAT PROGRAM. Do you see the issue now? Humans do stupid things from time to time, that much is a given even before pointing out Tide Pods, but you don't have all humans doing the same stupid thing that results in fatalities. With AI, if that problem exists, it will exist wholesale. THAT is the issue that you refuse to accept.

None of this matters, because that's not how you calculate risks. All that matters is the total number of accidents in two potential scenarios: the one we live in now, with 100% human drivers, and a hypothetical one with 100% automated cars. If the accident rate in the second case is lower you ban human drivers and replace them with AI. If the accident rate in the second case is higher then you work on improving AI before implementing it. It doesn't matter how the overall human-driver risk breaks down between particular humans because that's not the question you're asking. From a total deaths point of view there is zero difference between a binary distribution between high-risk and low-risk drivers and an even risk level among all drivers as long as the end result is the same number of deaths. 100 deaths is 100 deaths no matter how you get there, and it's still more than 95 deaths.

Now, it would in theory matter from a prevention point of view, if you want to do detailed studies about improving the safety of AI vs. programs to improve the safety of human drivers. A binary distribution would allow you to focus efforts on a small subset of drivers that cause the greatest risk, and potentially involve different risk-reduction strategies. But at this point, with so many years of accidents establishing such a solid record of human drivers taking risks and causing accidents, there is no reason to believe that the problem can be solved by trying to make people safer drivers. A certain number of people are simply going to be killed or injured in accidents involving human drivers.

I understand risk management explicitly. I use these principles at work both in the machine shop AND in the military. Variance is to be expected, as well as outliers. You refuse to accept that outliers are outliers while dealing with humans and refuse to accept the concept of reliably duplicated faults in AI that will pop up and need to be fixed. Yes, there ARE methods of correcting drivers. Our traffic laws approach that every day. Jammers in cars to prevent cell signal would be another, since teens and some young adults don't have the impulse control to put their phones down while driving, However, you won't accept that because of your particular political bend.

Let me rephrase your statement to better reflect that.


None of that matters because my communist dream of not working and having all my needs catered to by taking money from the wealthy can't come to fruition unless automation takes most if not all jobs.


Here's a thought from a corporation's point of view: If nobody is working and getting paid to make our products and others products, who will be BUYING our products? That is why automation will not replace the workforce: commerce will simply cease to exist. Well, I guess it'd have to go the USSR route and have the state take over production of goods. That's the dream in the end, isn't it? Same snake oil as the old burgoise/proletariat argument, just repackaged to be more palatable to the new generations?

The flaw in the Uber car programming is not in Google car programming.

They are different programs and the Google one is a lot better, though no doubt it has its own particular flaws.

Tony, you need to go read up on the Prisoner's Dilemna.

Kilkrazy wrote:The flaw in the Uber car programming is not in Google car programming.

They are different programs and the Google one is a lot better, though no doubt it has its own particular flaws.

Fair enough, but it shows an example of a system wide flaw. Can we agree on that? So ONLY Uber cars will misjudge and kill pedestrians. So they fix it. Got it. We stumble on one flaw that severe in one programming suite (sure there's a better term, but with the baby home from the hospital my time is finite) after implementation, and now we think there won't be another in other systems? That's how development works. That's how we find out that air bags become frag grenades in some instances. We learn and move on. My take is we need to keep doing controlled testing before we fling this tech out wholesale. Others don't agree.

skyth wrote:Tony, you need to go read up on the Prisoner's Dilemna.

I did, can you elaborate how a gaming concept has ANY bearing on the conversation?

Because you mentioned that businesses will not automate jobs because they need people to buy their products. Problem is each company has the incentive to automate but if everyone automates you run into issues...

 Just Tony wrote:
The flaw that allowed the AI car to kill that woman is in EVERY SINGLE AUTOMATED CAR USING THAT PROGRAM.

Did you see that the investigation turned up solid information that the automated driver people DISABLED the factory braking/safety programs??

There is a reason why increasing number of manufacturers are starting to do their own research and development on automated technology. They already have tons of driver assist capabilities, and there are significant issues when one company thinks it knows better than another, and turns their gak off because, reasons.

 Just Tony wrote:
Here's a thought from a corporation's point of view: If nobody is working and getting paid to make our products and others products, who will be BUYING our products? That is why automation will not replace the workforce: commerce will simply cease to exist.

You've made an assumption that each company makes its business decisions based on its need to preserve the market used by all companies. This is a junk assumption and it's going to produce some really broken analysis.

Read up on the Prisoner's Dilemma and The Tragedy of the Commons. It will teach you how each individual/company makes decisions based on their own personal interests, which can produce situations which are sub-optimal for the whole. For instance, if there's 1,000 companies and each of them is given the option to fully automate or keep using humans to manufacture, then each company is going to automate if that is cheaper. Sure it might kill demand overall, but each company is only responsible for 1,000 of total demand on average.


Automatically Appended Next Post:

 Just Tony wrote:
I did, can you elaborate how a gaming concept has ANY bearing on the conversation?

Gaming concept? Uh, game theory isn't about playing games. It's about using mathematical models to analyse individual decision making, and tweaking the rules in that model to try and make the individual's decision as optimal as possible for the overall group.

It is a huge part of economics. In this care your economic argument that companies won't automate because it will harm demand ignores the self-motivated actions of each individual company.

Yeah, like how Henry Ford didn't raise employee pay so they could afford the products he made, Or how other employers did the same because they saw that having no market generates revenue. No, wait, they DID do that. Despite Ford making advances that should have crippled his workforce.

Maybe I just need to rerereread the Prisoner's Dilemma, The Tragedy of the Commons, and throw in pretty much anything by Max Weber while I'm at it. Get my mind right.

 Just Tony wrote:
Yeah, like how Henry Ford didn't raise employee pay so they could afford the products he made, Or how other employers did the same because they saw that having no market generates revenue. No, wait, they DID do that. Despite Ford making advances that should have crippled his workforce.

That's an often quoted story that's actually total nonsense. Ford didn't raise wages to produce his own market. I mean how the feth would that work anyway, one company keeping an economy afloat with wages? What nonsense.

Anyhow, the truth is Ford increased wages to $5 a day because work on his factory lines was extremely demanding, and despite already paying a relatively high $2.50 a day he was still suffering extremely high turnover, after turnover was four months. This crippled productivity. So Ford raised wages to make people stay.

And if you still don't get it, Ford's labour force at the time was about 15,000 people. It was producing about 200,000 cars. In what kind of broken, nonsense maths do you think a workforce of 15,000 people are going to produce sufficient demand for a production run of 200,000?

Maybe I just need to rerereread the Prisoner's Dilemma, The Tragedy of the Commons, and throw in pretty much anything by Max Weber while I'm at it. Get my mind right.

Well you should know game theory isn't about games, for a starter.

 Just Tony wrote:
I understand risk management explicitly.

You clearly don't, because you're saying things that display a very poor understanding of the subject.

You refuse to accept that outliers are outliers while dealing with humans and refuse to accept the concept of reliably duplicated faults in AI that will pop up and need to be fixed.

No, I accept the concept of AI faults, they are just accounted for in the overall risk numbers. Again, it doesn't matter how you get to a final deaths per year number, only what that number is. If AI vehicles, despite having faults pop up and need fixing, kill/injure fewer people per year than the human drivers they replace then the AI is a success. You keep arguing that the AI will have faults and ignoring the fact that even a flawed AI that causes accidents can still be safer than human drivers, and implementing it can be a net improvement in safety.

Also, bad human drivers are not outliers. People driving drunk, while on their phones, etc, are common events. Drunk driving alone is ~10k deaths per year in the US and ~1 million arrests per year, or an average of a drunk driving death and 100 arrests every hour. AI vehicles would have to kill 10,000 people per year just to equal the deaths from drunk drivers, a problem that is 100% eliminated by removing human control over vehicles.

Yes, there ARE methods of correcting drivers. Our traffic laws approach that every day. Jammers in cars to prevent cell signal would be another, since teens and some young adults don't have the impulse control to put their phones down while driving, However, you won't accept that because of your particular political bend.

First of all, the idea that I won't accept these things because of my "particular political bend" is just laughable. You have no idea what my position on those ideas is, it's just convenient for you to assume I'm blindly opposed to your great ideas because of "politics".

Second, those ideas are nice in theory but the real world doesn't match your theory. It is indisputable fact that we are not currently implementing them, and the laws we do have (for example, against drunk driving) are not effective. And there does not appear to be any meaningful effort to make the laws you are suggesting in the foreseeable future. You can talk all you want about what could happen in theory, but the reality is that these things continue to kill people.

None of that matters because my communist dream of not working and having all my needs catered to by taking money from the wealthy can't come to fruition unless automation takes most if not all jobs.

It's not a dream, it's simple fact. Automation is improving. Capitalist desire for increased profits will continue to drive it forward and eliminate jobs. Eventually it will reach a point where communism/socialism is the only possible answer to the unemployment problem. Whether you or I want these things to happen is irrelevant. The technology is going to be developed, and the market pressure to implement it is extremely powerful.

Here's a thought from a corporation's point of view: If nobody is working and getting paid to make our products and others products, who will be BUYING our products? That is why automation will not replace the workforce: commerce will simply cease to exist. Well, I guess it'd have to go the USSR route and have the state take over production of goods. That's the dream in the end, isn't it? Same snake oil as the old burgoise/proletariat argument, just repackaged to be more palatable to the new generations?

Other people covered this, and they are correct. If I'm in charge of a company I don't give a about the economy as a whole. I care about the fact that automation can massively reduce my labor costs and put a ton of extra profit in my pocket. That's why companies have a long history of killing off jobs with automation and pocketing the extra profits, without any concern for what the now-unemployed people will do. And if I don't do it now my competition will, allowing them to undercut my prices and put me out of business. Automation is the only acceptable answer, as fast as the technology allows it to happen.

The only way to escape the automation trap is increased state control of business, making automation illegal and imposing minimum employee quotas that must be maintained. And you know, that sounds an awful lot like all that communism and planned economies stuff you claim to hate...


Automatically Appended Next Post:

 Just Tony wrote:
Maybe I just need to rerereread the Prisoner's Dilemma, The Tragedy of the Commons, and throw in pretty much anything by Max Weber while I'm at it. Get my mind right.

Add as much sarcasm as you want, but you absolutely do need to read those things. It's impossible to make any kind of informed comment on economics without at least a basic understanding of game theory. Even the most pro-capitalist economic theories depend on game theory principles, making your attempt to present them as some kind of leftist/communist fringe ideas pretty amusing.

 Peregrine wrote:
No, I accept the concept of AI faults, they are just accounted for in the overall risk numbers. Again, it doesn't matter how you get to a final deaths per year number, only what that number is. If AI vehicles, despite having faults pop up and need fixing, kill/injure fewer people per year than the human drivers they replace then the AI is a success. You keep arguing that the AI will have faults and ignoring the fact that even a flawed AI that causes accidents can still be safer than human drivers, and implementing it can be a net improvement in safety.

Also, bad human drivers are not outliers. People driving drunk, while on their phones, etc, are common events. Drunk driving alone is ~10k deaths per year in the US and ~1 million arrests per year, or an average of a drunk driving death and 100 arrests every hour. AI vehicles would have to kill 10,000 people per year just to equal the deaths from drunk drivers, a problem that is 100% eliminated by removing human control over vehicles.

To add to this: Google's driverless cars are already better than humans and they are still in the prototype stage for various reasons (I think thee was recently some new development that'll allow for drastic miniaturisation of LIDAR equipment). They probably could (if legislation were ready and people were willing to have clunky add-ons on their cars) sell kits that would save lives, at least in relatively clean conditions (I think regular detritus and rain is okay for their system but winter is still not optimal). If I remember correctly all accidents their autonomous cars were involved in were caused by humans and not the AI. From a year ago (http://bigthink.com/ideafeed/googles-self-driving-car-is-ridiculously-safe):

"We just got rear-ended again yesterday while stopped at a stoplight in Mountain View. That's two incidents just in the last week where a driver rear-ended us while we were completely stopped at a light! So that brings the tally to 13 minor fender-benders in more than 1.8 million miles of autonomous and manual driving — and still, not once was the self-driving car the cause of the accident.”

Some more: https://mashable.com/2012/08/07/google-driverless-cars-safer-than-you/?europe=true

Google announced Tuesday that its self-driving cars have completed 300,000 miles of test-drives, under a "wide range of conditions," all without any kind of accident. (The project has seen a few accidents in the past — but only with humans at the wheel.)

To put that into perspective, the average U.S. driver has one accident roughly every 165,000 miles. Here's how we got that figure: our average mileage per year is 16,550, according to the Federal Highway Administration; the average length of time we go between traffic accidents is 10 years, according to Allstate. (In particularly safe cities such as Fort Collins, Colo., that number can rise to 14 years — which is still no match for Google's 300,000 miles.)

And on the funny side (http://www.thedrive.com/sheetmetal/15023/autonomous-cars-are-getting-into-accidents-because-they-drive-too-well):

With self-driving test mules roaming around California, it’s been found that some of them are overly cautious. Human drivers are so unaccustomed to cars coming to a complete stop and obeying the speed limit that it’s resulting in a lot of low-speed accidents according to a Bloomberg report.

 Just Tony wrote:

 Peregrine wrote:

 BaronIveagh wrote:
That said, do you honestly thing that regulations for self driving cars will be nearly what they need to be when they hit the street compared to what they might be 40 years after the technology matures?

Again, you are making the wrong comparison. It doesn't matter how automated cars now compare to automated cars in the future. Of course the technology is going to improve, nobody is disagreeing with that idea. But the question that matters is how the current technology compares to the known poor safety of human drivers. Automated cars can have safety problems and kill people but still be an improvement over human drivers, and if that improvement exists they should become mandatory even if they aren't perfect yet.

You forgot to sling a few digs about cell phone use and the like. Here's the beef, and one you ignore. Not everyone texts while driving. I sincerely doubt that it hits 50% amongst drivers over the age of 21. The flaw that allowed the AI car to kill that woman is in EVERY SINGLE AUTOMATED CAR USING THAT PROGRAM. Do you see the issue now? Humans do stupid things from time to time, that much is a given even before pointing out Tide Pods, but you don't have all humans doing the same stupid thing that results in fatalities. With AI, if that problem exists, it will exist wholesale. THAT is the issue that you refuse to accept.

Before we blame that whole incident on the AI, we do need to remember the woman in question walked right out in front of a car doing 40 mph, and not at a crosswalk or stop sign. She must take a certain amount of responsibility for her own death.

And... I wouldn't be so sure that EVERY SINGLE AUTOMATED CAR USING THAT PROGRAM still has that flaw. I'm quite sure they have eliminated it in an update now that they have identified the flaw. In contrast, we've been trumpeting about how driving drunk kills people for over forty years that I am aware of... and drunk drivers STILL kill hundreds of people a year, if not thousands.

EDIT: Excuse me. TEN thousand. Per year. For at least forty years. We don't seem to be doing all that great a job fixing known flaws in human drivers....

 Vulcan wrote:
EDIT: Excuse me. TEN thousand. Per year. For at least forty years. We don't seem to be doing all that great a job fixing known flaws in human drivers....

This is the key element that a lot of people are refusing to get. If a human driver screws up and has a serious accident, then it is a learning experience for that human driver and no-one else. That one guy finds out that driving the speed limit in the wet on poor roads might be legal, but its also very stupid. Other drivers will have to have their own near death experiences to learn that lesson. Whereas an AI that doesn't fully account for those conditions and crashes will lead to an adjustment not just in that car but in all similar makes and models.

AI driving systems can fail forward in a way that human drivers haven't.

Another factor to consider- while it is true that a programming flaw like the one that caused the death of the woman was present in all of the cars- not all the self driving cars killed someone before the problem was identified and solved. You can bet that competition to create the safest AI program is huge- because companies will have to payout big time for lawsuits with any known defects that lead to injury or death.

This isn't some static system- every year the programming will get better, problems will be discovered and resolved, probably even optimized for different terrain, or states with the individual limits and such built in.

Compare that against humans, who, as a group, are about as safe as we're going to get as drivers. We're young, nervous and stupid, then pretty good for a while, and finally our reflexes go, our vision decreases and our pride keeps us driving longer than we're physically able to do safely.

I would love for this technology to become standardized, so my children need never learn how to drive, or get into the inevitable traffic incidents that occur in a lifetime.