I can't say I've every hired a car for vacation or dove a long road trip. Your fallacy is that everyone has an identical vacation in an identical area at the same time.
THAT is silly.
Due to school holidays, the majority of people take their vacations at the same time. This is not a fallacy, it's a fact. It's why vacation spots have a separate off-peak rate, because, well, "off-peak" literally means "not during peak demand". You appear to be making the somewhat stupid argument that there is no peak period for vacationing. Like I said, that's a stupid argument, and you should feel stupid for making it.
getting an expected smooth curve out of most of the independent trials implies some level of process control and precision,
I'm afraid it does nothing of the sort. Ever seen deliberately massaged data? It always follows the expected curve with a few outliers thrown in for legitimacy.
As you say, you're an engineer, so you're expecting data to be direct from a sensor with only non-human interference (EM, static, noise, etc). As a scientist my expectation is that the data has been massaged to fit the hypothesis.
(Engineers are nowhere near as skeptical of other humans as scientists are:-))
If I really want a roadtrip I can also hire the car for the roadtrip
This stupid argument comes up every time range is mentioned. You *won't* be able to hire a long-range car for your vacation because everyone else is also trying to hire a car for their vacation. If the rental companies keep enough cars for 90% of the population that only gets used twice a year they'll have to raise rates far beyond what you are prepared to pay.
It's a stupid argument, and you should feel silly for using it!
As an engineer, I'm used to noisy data so was still able to find the data useful.
[...]
Another obviously, completely incorrect reading, confirmed by a doctor-ordered test at another lab.
As a scientist (working with lots of engineers), I respectfully disagree that you are finding the data useful. You are only discarding the obviously incorrect values while keeping the non-obviously incorrect values. IOW, without a control you don't know which of the values (within the range you consider "valid") are correct and which are not.
This is Linux and LKML, nothing is done in private. All conversations and development is done in public. And I have yet to see Linus rant against "new unknown guy" unless said person have really really fucked up beyond repair (I somehow recall a situation some years ago where an individual was trying to get in a patch in order to get "Linux Kernel developer" in his CV so he submitted bad patch after bad patch until Linus lost it), he only rants at people who both can take it and that should know better.
I agree with everything you posted; I was just pointing out that the GP presented a false dichotomy: that the only options were "lollipops" or "profanity'. This is clearly not true. There are options besides treating someone like a special snowflake and handing them lollipops.
Calm and rational discussion should be the norm, absolutely. No one argues with that. But, when someone fucks up, you don't hand them a lollipop and reassuringly pet their head.
Those aren't the only two options. You are acting as if the only two options are "lollipops and pettings" and "berate them in public". You can berate them in private, and firmly (but politely) mark them down in public.
No one here advocated treating them with kid gloves.
Except all accidents are based on per mile. And automated cars have been proven repeatedly to be much safer per mile than almost any grouping of humans.
They have never been proven to be safer. They have never even demonstrated an ability to actually self-drive on real roads. There is no a single demonstration of such. All "demos" have had a human in the drivers seat to correct the car as it drove.
There's no need: C++ has modtly caught up to Ada and has far better IDEs and compilers. All you need to do is stop using the "C" part and you're golden.
C++ isn't C. If I open Visual C++ and type this code I'll get an exception when I try to run it:
std::vector a;
a[-1] = 0x666;
So next time you hear Linus Torvalds rant against C++ tell him to fuck off and stop being an idiot.
That's a disingenuous example. C doesn't have vectors. If you open Visual C++ and type the following code in you won't get any exceptions:
int a[12];
a[-1] = 0x666;
Which, by the way, is exactly the same result you will get in C. What you may or may not have noticed in your self-professed long career in development is that C and C++ get used for different problems. If your problem doesn't allow for much of the C++ scaffolding then you're programming in C, like it or not, in which case you'll write your own vector library in C (or use one of the many platform-independent ones).
Honestly, if you wanted to maximize population expansion rate and you were hand-selecting the crew, you'd send 100% female and cryopreserved female embryos. You'd choose women with small stature to maximize how many you can send / keep alive with a give payload mass, and ideally from families/cultures that tend to have large numbers of children starting at a young age.
In practice, of course, there are other factors beyond maximizing reproduction. Particularly if the people going are paying customers rather than people being selected by some external organization.
I'd think that, considering the risks, a single failure in power and all the frozen embryos will die. The advantage of sending adults who produce their own gametes is that, as long as a single couple survives, he will produce unlimited gametes for the duration of his lifespan (barring exposure to radioactivity or solar flares or whatever the hell exists in the harsh environment) and she can produce a single child a year.
Whats more, using adults as containers for gametes gives you automatic protection and defense of those gametes. The freezer containing embryos will not even be able to identify danger, much less retreat from it.
Why would you send 50% female? To ensure success you'd want to send 2/3 female (or more). If you're limited in the number of people you can send (and he is), then you'd want to send as many young females as possible, with a few young men to impregnate them.
She just can't take it when people hold an opposing opinion. What a fucking intolerant cunt.
The proper term for that is a bigot, by definition. And Barbara is undeniably a bigot.
Yeah... Did she get into an argument with you where she resorts to saying that you're the type of man women never want to sleep with?
I got that repeatedly. I eventually had to point out that a) I've been married multiple times (so I've slept with more than one female), and b) She can continue ranting about how women must hate me but I had to go because my wife was waiting to give me my bi-weekly BJ:-)
Yeah I wouldn't worry about this - a court already found that "[the shooter[ had a right to shoot at the drone". how is the drone operator going to interpret the previous ruling? That the shooter can shoot at, but he must miss?
"This in turn allows the compiler to find other errors such as variables that are never initialized."
Maybe a code smell but not an error, but surely this is more a function of an IDE rather than the compiler?
Once you have a list of memory locations that are never initialised it's trivial to check if those locations are ever used, which is why it's handy to have a list of variables that are never initialised. For example, you *definitely* want to catch the following error:
var x; ...(hundreds of lines later)... if (x>0) {... }
AIUI, in Javascript that will generate a run-time exception (if it doesn't, it *should*!), while in typescript that will generate an error at the time of compilation, before the code ever even runs.
Yesterday, researchers on behalf of Microsoft said they will "solve" cancer within the next 10 years by treating it like a computer virus that invades and corrupts the body's cells. Today, Facebook CEO Mark Zuckerberg and his wife Priscilla Chan announced a $3 billion initiative to "cure all diseases."
"I see how it is. Fine. I, Jeff Bezos, pledge an end to all human suffering by sometime in the next six months."
"Does no one else think cars + computers + network connectivity = bad?"
Nope. Tesla was able to patch all their cars quickly, without asking drivers to come in to get serviced.
???
The patch would not have been needed had the connectivity not existed.
"Luckily, this problem that would not have existed without network connectivity was solved by using the network connectivity." Circular reasoning at its finest, folks. There would have been no patch if there was no network connectivity.
Or would that depend on age? Cleaning cells at age 12 would have little effect, as mutations are fewer, but in 80+, it would result in death,
That's because cancer is a side-effect of living. When you can figure out how to prevent mutations in cell division, you would have cured cancer, but that same tech also gives immortality.
No your just being obtuse for arguments sake. A driver isn't a driver if they aren't driving. Calling them a driver is to fulfill a current regulatory requirement to have a driver... so you are arguing a chicken and an egg.
No. They have an out on that "regulatory hurdle" of having a driver in the seat - they can demonstrate the car in a variety of real-world but temporarily closed off areas. The reason you haven't seen many demonstrations of such (no precautionary driver) is because the software is not there yet. The claims have gotten louder in the last decade, but the claims are still the same - driver to provide regular corrections to the autonomous software.
They had to have someone called a driver even if they weren't driving because they required a driver.
That Tesla driver I say reading a book at the wheel certainly wasn't "driving" in any meaningful sense of the word... he was merely at the wheel... ready, sort of. Yes, that is just highway driving, but it is fully autonomous highway driving from a technical standpoint... just not a legal one yet.
Come now! You are making claims that the manufacturer disputes. It's purely a technical issue - the Tesla system is capable of doing exactly what was being done in the 90's. No more.
There is a big difference between computational power now and even just 5 years ago. With more computer power you get better image recognition in a variety of conditions. What took 8 seconds to identify a road sign on a laptop 5 or ten years ago takes 20 ms.
Sure you could autonomously drive decades ago with a truck full of computer hardware or do simple things with proximity sensors, but the envelope of capabilities is clearly getting a lot bigger and the affordability has come down well under the $100k range for a fully autonomous capable platform.
Being more affordable isn't going to complete that last bit that makes a car fully autonomous. Having more computational power isn't going to solve intractable problems, and being able to find the 10e100th prime doesn't mean that you will be able to find the (10e100 + 1)th prime in time for the heat death of the universe. Writing an algorithm that works 98% of each day doesn't mean you're almost at a 100% solution. That measurement only works for engineering, not for mathematics.
If the software problem was that easy to solve, considering the immense payoff, it would have been solved by now. Bright minds have worked on it for decades; ask them how well it is going. Don't ask the people who spent the last few years designing circuits and throwing hardware at the problem in the hope of brute-forcing a solution.
Personally, I feel that the people most likely to crack the software problem are going to be a bunch of abstract mathematicians working on something unrelated, not a bunch of engineers working incrementally towards a solution, in the same way that it was a mathematician who gave us the computer, not engineers who designed and build machines. The neural-net approach, as well, isn't suitable due to the opacity of the solution.
We need smarter algorithms instead of faster hardware, and by all accounts we don't have it.
2004 Nobody won the DARPA grand challenge... no car completed the course.
2005 5 vehicles completed the course.
2007 they switched to an urban course having to obey the rules of the road and six teams finished the course.
And prior to 2004 there were already cars being demonstrated as autonomous with minimal human input. Poster above linked to one from the 90's.
That is rapid progress.
From 2007 to 2016 we have seen pretty steady progress with commercially available features for things like automatic parking, automated braking and collision avoidance, widespread use of GPS navigation (via smartphones and built-in) and more recently the fully autonomous highway driving from Tesla (yes I've seen the people reading books while "at the wheel").
Tesla themselves say it is not autonomous. Automated braking was demonstrated in the 70's on a Volvo. You list it as some sort of breakthrough.
And Google has been pretty open in their fully autonomous car project with two different cars one based on an off the shelf lexus and another custom built electric vehicle: 1.5 million miles driven and "currently out on the streets of Mountain View, CA, Austin, TX, Kirkland, WA and Metro Phoenix, AZ"
And we are seeing Uber's autonomous efforts play out in Pittsburgh. Multiple companies, multiple projects, multiple on-street implementations that are getting better and better.
Firstly, Uber themselves say that the drivers in their autonomous cars will remain there for the foreseeable future. Their words, not mine. Secondly, you yourself pointed out that they are at the same place they were 10 years ago - that's not "getting better". There's been incremental improvements, sure, there's been more widespread takeup of the existing technologies, sure... but the state-of-the-art today in autonomous cars is basically the same as ten years ago. The cars are mostly autonomous.... up until they aren't! There's no new tech. The improvements have been tiny and incremental, and not enough to replace a human driver yet.
What people don't get is that we've seen this movie before, in the field of AI. The AI titans were 99% there with regard to thinking machines. Turned out that that 1% was unattainable. Same deal with SDC software: they may have checked off 99 out of 100 requirements for the software, but that remaining requirement (requiring corrections from an alert human at the wheel) may or may not get there.
The AI winter that followed the "we're 99% there!" AI boom was painful as many researchers had to admit that they had an intractable problem on their hands. We are currently seeing the same thing in autonomous cars: everyone is so certain that they're almost there and the remaining problems will be easy. Well, if the remaining problem was that easy, it would have been solved when it first came up in the 90's.
The problem isn't the hardware (GPS, Vision capture, distance sensor, lane detection, etc), it's the software. No company has demonstrated software-controlled cars that can function without regular human input.
And this problem isn't an easy one, which is why it hasn't been solved. The only improvements you can list came from better and more accessible hardware.
A few of your points are somewhat absurd. Having an observer in the car is irrelevant to whether the car is actually driving autonomously or not...
Those aren't observers, as you point out below...
Having someone ready to take over if the car fails is a precaution.
You are quite right - those drivers are precautions. This makes SDCs look even worse, because, as you pointed out in your previous response, we've had that for at least a decade. For a decade (or longer) we've had self-driving cars with precautionary drivers. What is happening now with Uber and the rest is, as you pointed out NOTHING NEW!
These vehicles are being driven for hundreds of thousands of miles on a variety of roads and a variety of conditions.
No, they aren't. They are being driven hundreds of thousands of miles on well-demarcated roads with occasional input from the precautionary driver.
Heck, if I recall, the DARPA grand challenge was partially on a dirt road and that was ten years ago.
That is exactly my point - there has been very little, if any, progress on self-driving cars in the last decade. If the best that can be done now is what Tesla offers, you are correct in noting that we've had this for over a decade. I'm simply noting the lack of progress. Remember Uber's press release - "The engineer in the drivers seat will remain there for the foreseeable future.".
Once again I must point out that software is deceptive in estimating. For common business-logic software which has had some 40 years of refinements, we still adhere to "the first 90% of the project takes 90% of the time, the other 10% takes another 90% of the time". This is for well-understood and previously built software!. Imagine what it is like for the stuff that has never succeeded before. You'll never know how far you are from the end.
With software, just because you implemented 9 out of 10 requirements does not mean that the last requirement is at all possible. Your last requirement might just be intractable, moreso if you're going the neural net path (hehe) as there is a lot more unknowns no matter what you trained the net on.
This is one of the things needed to get this technology legal and on the road.
Before getting this technology legal and on the road, perhaps we should focus on getting this technology? For the last five years I've been hearing that "Self-driving cars are here already", but sadly they aren't.
Okay now they are here.
Where? I see driver-assist cars, but no self-driving cars.
You mean available to purchase by consumers?
No, I meant "available at all".
Okay, not yet. But Uber is rolling out commercial service using self driving cars right now
No, they aren't - there's a human in the drivers seat.
and multiple companies apparently have fully autonomous vehicles on the public roads now.
Wrong again - no company has demonstrated a self-driving car in anything other than perfect conditions.
And acedemic/research teams have had fully self driving cars for at least ten years.
Yeah, right - and these decade-old technology is soooo undesirable that neither Uber, nor Tesla, nor Lyft offered them money for it, right? It's just sitting there, in the university
At least the Uber example has to be considered as commercial availability since this is one of the ways companies will offer self driving cars to the public, on a per trip basis. They are here.
No, they are not. Word of their imminent arrival is here. Look around - all self-driving cars need a driver in the seat. Academia has been working on this problem since at least the 90's, and they have yet to make progress. They've got a whole bunch of new sensors, but they still don't have the software ready yet.
The problem is that software is deceptive in it's complexity. Hardware is simple to gauge complexity. A hardware system that is 90% of the way there is easy to look at and figure out an ETA. A software system can't even be measured to see if it is 90% there.
For at least the last 20 years self-driving software systems were "90% there". That last 10% might just be impossible, or hard, or easy, but you can't tell, especially if you're using a neural network ('cos that is opaque even to the developers). The state of self-driving cars has not changed in the last 10 years. The announcements have.
Slashdot Mirror
I can't say I've every hired a car for vacation or dove a long road trip. Your fallacy is that everyone has an identical vacation in an identical area at the same time.
THAT is silly.
Due to school holidays, the majority of people take their vacations at the same time. This is not a fallacy, it's a fact. It's why vacation spots have a separate off-peak rate, because, well, "off-peak" literally means "not during peak demand". You appear to be making the somewhat stupid argument that there is no peak period for vacationing. Like I said, that's a stupid argument, and you should feel stupid for making it.
getting an expected smooth curve out of most of the independent trials implies some level of process control and precision,
I'm afraid it does nothing of the sort. Ever seen deliberately massaged data? It always follows the expected curve with a few outliers thrown in for legitimacy.
As you say, you're an engineer, so you're expecting data to be direct from a sensor with only non-human interference (EM, static, noise, etc). As a scientist my expectation is that the data has been massaged to fit the hypothesis.
(Engineers are nowhere near as skeptical of other humans as scientists are :-))
If I really want a roadtrip I can also hire the car for the roadtrip
This stupid argument comes up every time range is mentioned. You *won't* be able to hire a long-range car for your vacation because everyone else is also trying to hire a car for their vacation. If the rental companies keep enough cars for 90% of the population that only gets used twice a year they'll have to raise rates far beyond what you are prepared to pay.
It's a stupid argument, and you should feel silly for using it!
And yet flood legends are found all across the world (https://en.wikipedia.org/wiki/List_of_flood_myths)
That's because floods are found all across the world, not because there was a single global flood.
As an engineer, I'm used to noisy data so was still able to find the data useful.
[...]
Another obviously, completely incorrect reading, confirmed by a doctor-ordered test at another lab.
As a scientist (working with lots of engineers), I respectfully disagree that you are finding the data useful. You are only discarding the obviously incorrect values while keeping the non-obviously incorrect values. IOW, without a control you don't know which of the values (within the range you consider "valid") are correct and which are not.
This is Linux and LKML, nothing is done in private. All conversations and development is done in public. And I have yet to see Linus rant against "new unknown guy" unless said person have really really fucked up beyond repair (I somehow recall a situation some years ago where an individual was trying to get in a patch in order to get "Linux Kernel developer" in his CV so he submitted bad patch after bad patch until Linus lost it), he only rants at people who both can take it and that should know better.
I agree with everything you posted; I was just pointing out that the GP presented a false dichotomy: that the only options were "lollipops" or "profanity'. This is clearly not true. There are options besides treating someone like a special snowflake and handing them lollipops.
Calm and rational discussion should be the norm, absolutely. No one argues with that. But, when someone fucks up, you don't hand them a lollipop and reassuringly pet their head.
Those aren't the only two options. You are acting as if the only two options are "lollipops and pettings" and "berate them in public". You can berate them in private, and firmly (but politely) mark them down in public.
No one here advocated treating them with kid gloves.
Except all accidents are based on per mile. And automated cars have been proven repeatedly to be much safer per mile than almost any grouping of humans.
They have never been proven to be safer. They have never even demonstrated an ability to actually self-drive on real roads. There is no a single demonstration of such. All "demos" have had a human in the drivers seat to correct the car as it drove.
They're ready, once they're as safe as humans on the road. Which happened several years ago.
They haven't happened *yet*, nevermind "years ago".
Why are the people most likely to complain about "SJW" also the most likely to drag up this sort of topic at every available opportunity?>
--
I started watching "Tropes Versus Women in Video Games" recently due to the persistent lies about fraud. It's terrific!
Why are you complaining? With your sig you bring up this sort of topic in every post you make. That's more than slightly hypocritical.
There's no need: C++ has modtly caught up to Ada and has far better IDEs and compilers. All you need to do is stop using the "C" part and you're golden.
C++ isn't C. If I open Visual C++ and type this code I'll get an exception when I try to run it: std::vector a; a[-1] = 0x666;
So next time you hear Linus Torvalds rant against C++ tell him to fuck off and stop being an idiot.
That's a disingenuous example. C doesn't have vectors. If you open Visual C++ and type the following code in you won't get any exceptions:
int a[12];
a[-1] = 0x666;
Which, by the way, is exactly the same result you will get in C. What you may or may not have noticed in your self-professed long career in development is that C and C++ get used for different problems. If your problem doesn't allow for much of the C++ scaffolding then you're programming in C, like it or not, in which case you'll write your own vector library in C (or use one of the many platform-independent ones).
Honestly, if you wanted to maximize population expansion rate and you were hand-selecting the crew, you'd send 100% female and cryopreserved female embryos. You'd choose women with small stature to maximize how many you can send / keep alive with a give payload mass, and ideally from families/cultures that tend to have large numbers of children starting at a young age.
In practice, of course, there are other factors beyond maximizing reproduction. Particularly if the people going are paying customers rather than people being selected by some external organization.
I'd think that, considering the risks, a single failure in power and all the frozen embryos will die. The advantage of sending adults who produce their own gametes is that, as long as a single couple survives, he will produce unlimited gametes for the duration of his lifespan (barring exposure to radioactivity or solar flares or whatever the hell exists in the harsh environment) and she can produce a single child a year.
Whats more, using adults as containers for gametes gives you automatic protection and defense of those gametes. The freezer containing embryos will not even be able to identify danger, much less retreat from it.
Why would you send 50% female? To ensure success you'd want to send 2/3 female (or more). If you're limited in the number of people you can send (and he is), then you'd want to send as many young females as possible, with a few young men to impregnate them.
She just can't take it when people hold an opposing opinion. What a fucking intolerant cunt.
The proper term for that is a bigot, by definition. And Barbara is undeniably a bigot.
Yeah... Did she get into an argument with you where she resorts to saying that you're the type of man women never want to sleep with?
I got that repeatedly. I eventually had to point out that a) I've been married multiple times (so I've slept with more than one female), and b) She can continue ranting about how women must hate me but I had to go because my wife was waiting to give me my bi-weekly BJ :-)
For some reason, that just pissed her off more :-)
Yeah I wouldn't worry about this - a court already found that "[the shooter[ had a right to shoot at the drone". how is the drone operator going to interpret the previous ruling? That the shooter can shoot at, but he must miss?
"This in turn allows the compiler to find other errors such as variables that are never initialized."
Maybe a code smell but not an error, but surely this is more a function of an IDE rather than the compiler?
Once you have a list of memory locations that are never initialised it's trivial to check if those locations are ever used, which is why it's handy to have a list of variables that are never initialised. For example, you *definitely* want to catch the following error:
AIUI, in Javascript that will generate a run-time exception (if it doesn't, it *should*!), while in typescript that will generate an error at the time of compilation, before the code ever even runs.
Yesterday, researchers on behalf of Microsoft said they will "solve" cancer within the next 10 years by treating it like a computer virus that invades and corrupts the body's cells. Today, Facebook CEO Mark Zuckerberg and his wife Priscilla Chan announced a $3 billion initiative to "cure all diseases."
"I see how it is. Fine. I, Jeff Bezos, pledge an end to all human suffering by sometime in the next six months."
[fineprint]Only for Prime Members[/fineprint]
"Does no one else think cars + computers + network connectivity = bad?" Nope. Tesla was able to patch all their cars quickly, without asking drivers to come in to get serviced.
???
The patch would not have been needed had the connectivity not existed.
"Luckily, this problem that would not have existed without network connectivity was solved by using the network connectivity." Circular reasoning at its finest, folks. There would have been no patch if there was no network connectivity.
I barely lose weight during the winter
Bears put on weight for the winter not lose it.
Can you tell the difference between "doing something for winter" and "doing something during winter"?
Or would that depend on age? Cleaning cells at age 12 would have little effect, as mutations are fewer, but in 80+, it would result in death,
That's because cancer is a side-effect of living. When you can figure out how to prevent mutations in cell division, you would have cured cancer, but that same tech also gives immortality.
If your definition of a human is a retarded 4-year-old that can be trained to name colors with 75% accuracy, yes.
We're not there, we're not even close; "machine learning" is just the new buzzword in town, rising from the ashes of "big data".
Big data itself was a phoenix from the ashes of AI.
No your just being obtuse for arguments sake. A driver isn't a driver if they aren't driving. Calling them a driver is to fulfill a current regulatory requirement to have a driver... so you are arguing a chicken and an egg.
No. They have an out on that "regulatory hurdle" of having a driver in the seat - they can demonstrate the car in a variety of real-world but temporarily closed off areas. The reason you haven't seen many demonstrations of such (no precautionary driver) is because the software is not there yet. The claims have gotten louder in the last decade, but the claims are still the same - driver to provide regular corrections to the autonomous software.
They had to have someone called a driver even if they weren't driving because they required a driver.
That Tesla driver I say reading a book at the wheel certainly wasn't "driving" in any meaningful sense of the word... he was merely at the wheel... ready, sort of. Yes, that is just highway driving, but it is fully autonomous highway driving from a technical standpoint... just not a legal one yet.
Come now! You are making claims that the manufacturer disputes. It's purely a technical issue - the Tesla system is capable of doing exactly what was being done in the 90's. No more.
There is a big difference between computational power now and even just 5 years ago. With more computer power you get better image recognition in a variety of conditions. What took 8 seconds to identify a road sign on a laptop 5 or ten years ago takes 20 ms.
Sure you could autonomously drive decades ago with a truck full of computer hardware or do simple things with proximity sensors, but the envelope of capabilities is clearly getting a lot bigger and the affordability has come down well under the $100k range for a fully autonomous capable platform.
Being more affordable isn't going to complete that last bit that makes a car fully autonomous. Having more computational power isn't going to solve intractable problems, and being able to find the 10e100th prime doesn't mean that you will be able to find the (10e100 + 1)th prime in time for the heat death of the universe. Writing an algorithm that works 98% of each day doesn't mean you're almost at a 100% solution. That measurement only works for engineering, not for mathematics.
If the software problem was that easy to solve, considering the immense payoff, it would have been solved by now. Bright minds have worked on it for decades; ask them how well it is going. Don't ask the people who spent the last few years designing circuits and throwing hardware at the problem in the hope of brute-forcing a solution.
Personally, I feel that the people most likely to crack the software problem are going to be a bunch of abstract mathematicians working on something unrelated, not a bunch of engineers working incrementally towards a solution, in the same way that it was a mathematician who gave us the computer, not engineers who designed and build machines. The neural-net approach, as well, isn't suitable due to the opacity of the solution.
We need smarter algorithms instead of faster hardware, and by all accounts we don't have it.
2004 Nobody won the DARPA grand challenge... no car completed the course.
2005 5 vehicles completed the course.
2007 they switched to an urban course having to obey the rules of the road and six teams finished the course.
And prior to 2004 there were already cars being demonstrated as autonomous with minimal human input. Poster above linked to one from the 90's.
That is rapid progress.
From 2007 to 2016 we have seen pretty steady progress with commercially available features for things like automatic parking, automated braking and collision avoidance, widespread use of GPS navigation (via smartphones and built-in) and more recently the fully autonomous highway driving from Tesla (yes I've seen the people reading books while "at the wheel").
Tesla themselves say it is not autonomous. Automated braking was demonstrated in the 70's on a Volvo. You list it as some sort of breakthrough.
And Google has been pretty open in their fully autonomous car project with two different cars one based on an off the shelf lexus and another custom built electric vehicle: 1.5 million miles driven and "currently out on the streets of Mountain View, CA, Austin, TX, Kirkland, WA and Metro Phoenix, AZ"
And we are seeing Uber's autonomous efforts play out in Pittsburgh. Multiple companies, multiple projects, multiple on-street implementations that are getting better and better.
Firstly, Uber themselves say that the drivers in their autonomous cars will remain there for the foreseeable future. Their words, not mine. Secondly, you yourself pointed out that they are at the same place they were 10 years ago - that's not "getting better". There's been incremental improvements, sure, there's been more widespread takeup of the existing technologies, sure... but the state-of-the-art today in autonomous cars is basically the same as ten years ago. The cars are mostly autonomous.... up until they aren't! There's no new tech. The improvements have been tiny and incremental, and not enough to replace a human driver yet.
What people don't get is that we've seen this movie before, in the field of AI. The AI titans were 99% there with regard to thinking machines. Turned out that that 1% was unattainable. Same deal with SDC software: they may have checked off 99 out of 100 requirements for the software, but that remaining requirement (requiring corrections from an alert human at the wheel) may or may not get there.
The AI winter that followed the "we're 99% there!" AI boom was painful as many researchers had to admit that they had an intractable problem on their hands. We are currently seeing the same thing in autonomous cars: everyone is so certain that they're almost there and the remaining problems will be easy. Well, if the remaining problem was that easy, it would have been solved when it first came up in the 90's.
The problem isn't the hardware (GPS, Vision capture, distance sensor, lane detection, etc), it's the software. No company has demonstrated software-controlled cars that can function without regular human input.
And this problem isn't an easy one, which is why it hasn't been solved. The only improvements you can list came from better and more accessible hardware.
A few of your points are somewhat absurd. Having an observer in the car is irrelevant to whether the car is actually driving autonomously or not...
Those aren't observers, as you point out below...
Having someone ready to take over if the car fails is a precaution.
You are quite right - those drivers are precautions. This makes SDCs look even worse, because, as you pointed out in your previous response, we've had that for at least a decade. For a decade (or longer) we've had self-driving cars with precautionary drivers. What is happening now with Uber and the rest is, as you pointed out NOTHING NEW!
These vehicles are being driven for hundreds of thousands of miles on a variety of roads and a variety of conditions.
No, they aren't. They are being driven hundreds of thousands of miles on well-demarcated roads with occasional input from the precautionary driver.
Heck, if I recall, the DARPA grand challenge was partially on a dirt road and that was ten years ago.
That is exactly my point - there has been very little, if any, progress on self-driving cars in the last decade. If the best that can be done now is what Tesla offers, you are correct in noting that we've had this for over a decade. I'm simply noting the lack of progress. Remember Uber's press release - "The engineer in the drivers seat will remain there for the foreseeable future.".
Once again I must point out that software is deceptive in estimating. For common business-logic software which has had some 40 years of refinements, we still adhere to "the first 90% of the project takes 90% of the time, the other 10% takes another 90% of the time". This is for well-understood and previously built software!. Imagine what it is like for the stuff that has never succeeded before. You'll never know how far you are from the end.
With software, just because you implemented 9 out of 10 requirements does not mean that the last requirement is at all possible. Your last requirement might just be intractable, moreso if you're going the neural net path (hehe) as there is a lot more unknowns no matter what you trained the net on.
This is one of the things needed to get this technology legal and on the road.
Before getting this technology legal and on the road, perhaps we should focus on getting this technology? For the last five years I've been hearing that "Self-driving cars are here already", but sadly they aren't.
Okay now they are here.
Where? I see driver-assist cars, but no self-driving cars.
You mean available to purchase by consumers?
No, I meant "available at all".
Okay, not yet. But Uber is rolling out commercial service using self driving cars right now
No, they aren't - there's a human in the drivers seat.
and multiple companies apparently have fully autonomous vehicles on the public roads now.
Wrong again - no company has demonstrated a self-driving car in anything other than perfect conditions.
And acedemic/research teams have had fully self driving cars for at least ten years.
Yeah, right - and these decade-old technology is soooo undesirable that neither Uber, nor Tesla, nor Lyft offered them money for it, right? It's just sitting there, in the university
At least the Uber example has to be considered as commercial availability since this is one of the ways companies will offer self driving cars to the public, on a per trip basis. They are here.
No, they are not. Word of their imminent arrival is here. Look around - all self-driving cars need a driver in the seat. Academia has been working on this problem since at least the 90's, and they have yet to make progress. They've got a whole bunch of new sensors, but they still don't have the software ready yet.
The problem is that software is deceptive in it's complexity. Hardware is simple to gauge complexity. A hardware system that is 90% of the way there is easy to look at and figure out an ETA. A software system can't even be measured to see if it is 90% there.
For at least the last 20 years self-driving software systems were "90% there". That last 10% might just be impossible, or hard, or easy, but you can't tell, especially if you're using a neural network ('cos that is opaque even to the developers). The state of self-driving cars has not changed in the last 10 years. The announcements have.