I agree with you 100 percent. DAS (driver assist systems) are getting to be very common. I've worked at a couple companies that produce them, though I did not work on those products. I rode on the highway in a car outfitted with prototype lane keeping, and it was not quite as steady as I'd hoped. They also had an option that would apply a light force to keep you in the lane but could be easily overcome (and I think went away if you used a turn signal). It felt like a sort of speed bump between lanes. It wasn't trying to be a really complex system, and probably has some real world value - in reasonable situations I'd like to be able to use both hands to eat some food from the drive through, while keeping my eyes on things.
I've also seen video (circa 2004) of an autonomous car driving 100kph down a winding dirt road, staying on its side of the road and automatically stopping. But again, the engineers wanted to do so much more with it, but the rational guys in safety would not allow it.
I also competed in the AUVS autonomous ground vehicle competition in 1994. I wrote a lane follower by taking 20-30 lines of pixels off an NTSC frame grabber on an Intel 486. The core algorithm was on the order of 100 LoC, and we took second place. Super simple algorithm, hardly intelligent.
There is a huge difference between a PID controller maintaining position in the lane with radar assisted cruise control, and being a fully autonomous vehicle fit of unaided driving on public roads. There is a whole range of system capabilities, and the public has no idea what's in any given car. Comparing highway fatalities/mile to national statistics covering all roads and conditions is bullshit and a certain company that's recently killed some people knows it.
But hey, we all want to be able to read a book on the road or have truckers take a nap on long hauls, so let's keep deluding ourselves that this stuff will be ready for prime time in the next couple years.
How is this relevant to self driving cars in 2016 and the following decades?
Even as a human, I get confused while driving on this stretch. The old lane markings aren't quite faded out enough to easily tell them apart. As I'm driving, I see every other car slowing down and "dead reckoning," too.
How will the run-of-the-mill self-driving car's lane following handle this corner case? If it's a Canny edge detector (like most self-driving cars), it's definitely going to fail and find false positives.
The algorithms for lane following are quite a bit more sophisticated than that though. Here's a paper from 2008, and there are other approaches as well.
http://www.vision.caltech.edu/malaa/publications/aly08realti...
Take a look at Figure 12 for an exemplary look at its failure cases -- cases that are much less forgiving than the aforementioned Palo Alto stretch 101 lane markings.
(But in this case, I'll add that HOG doesn't solve the problem of overlapping old/new road markings)
What if I drive 5x better and safer than the "average driver"? Doesn't that mean that these cars would now put me in more danger than I would normally be in with my own manual car?
Self-driving cars should be 10x better than the best (as in safest by number of incidents, say, per decade) drivers on Earth. Then I could begin trusting them ... if only there wasn't that pesky security issue, too, with self-driving cars getting hacked and then driving you off the highway at 70mph and killing you.
We see Teslas getting remotely hacked. If that happens with Tesla, a Silicon Valley car company with a lot of hired top tech talent, how can we ever trust a company like GM or Volkswagen to build safe self-driving cars?
Then don't use the system?
You are giving thresholds - 50% to 2x better, 10x better. Before what? Before we ban manual driving? "10x better than the best driver" might make sense. Before we allow it on the road? Why not "better than the worst drivers we allow on the road"?
The most relevant point for Musk is probably "before the marketing works out". The market of people who consider themselves "substantially below average" drivers is much smaller than the market of people who consider themselves "not all that much better than average".
Because you are unlikely always that.
Are you tired? Do you have a cold? Are you hungry? Are you playing with the radio? Are you having a conversation with a passenger?
The huge advantage that self-driving cars will have is integrated systemic control from multiple sensor sources and multiple transducer outputs--and they will never get distracted.
Human drivers have mostly 1 input: vision. Human drivers have mostly two outputs: hand and foot.
The lag in that system from input to output is >100ms even if you are perfect (I will grant you an exception if your name is Michael Schumacher).
Automated systems will have far more inputs. Multiple vision sources. Multiple radars. Temperature readings. RPM readings. etc.
Automated systems have far more outputs. Steering is the same. Brakes on individual tires is the big starter. With electric, individual suspension becomes feasible. Transmissions change form and become electrically controlled.
And the lag in the system is probably <10ms. That's huge. That's the difference in reacting in roughly 3 meters (2.7 meters) vs 1/3 of a meter at 100kph (9 feet vs 1 foot roughly for 60mph).
Finally, automated systems will take subtle preventative actions all the time. They will "see" the child on bicycle on the sidewalk (the idiot detector as Google called it) and "know" that they need to find him again when they attempt to turn right or slow down significantly. They will see the dude on a skateboard and know that they need to give him an extra couple feet if they can in order to deal with him falling. I know very few drivers that are that diligent.
Why wouldn't you want to make them better?
How do you know that you're actually 5x better than the average driver and not just deluding yourself?
I don't see why you give wetware more leeway than hardware.
My point is that basic lane following can be done quite trivially, but it's nowhere near something one should entrust with the safety of human beings. Your question implies that you think all modern systems being deployed involve some really sophisticated AI. You don't know that.
Nope. LIDAR is used to determine geometry when a camera system can't tell WTF it's looking at.
Yes, that would be a communication problem of mine. It's something I'm working on ;-) But I would counter that LIDAR isn't going to fully make up for the problems of the camera based system with the result being a simple sounding statement like "if they're using LIDAR it's not good enough", or grossly oversimplified to "LIDAR sucks".
I will admit that LIDAR could conceivably enhance a really good vision system by offering better depth perception, but I don't believe that's why people are using it today.
This stuff is limited right now but the sooner these (actually stage 4 autonomous) cars starts hitting the road and being used in real life, the faster the tech will improve. I expect some dramatic improvements once Google car quality sensors are combined with full-data collection on hundreds of thousands of daily rides.
Of course, it won't become totally autonomous until the vast majority of the cars on the road are using assisted driving and potentially interfacing/coordinating with each other. At that point the speeds can be increased and the driver can be disconnected from the backup driver role.
With kits already coming out with the idea of 'back-porting' old cars to use new automated driving tech, I don't think it's a huge barrier to adoption. It won't require everyone having to go out to buy new mid-level cars to participate. People can purchased insurance (and personal safety) incentivized kits to make their cars automated.
[1] Comma.ai https://www.youtube.com/watch?v=KTrgRYa2wbI and technical talk: https://www.youtube.com/watch?v=Hxoke1lDJ9w
The level of arrogance in that is pretty staggering. Remember, I mentioned a system driving on (unmarked) dirt roads at highway speeds 10 years ago. I'm sorry, but advanced tech does exist outside of California.
And tech companies and AI experts also exist outside of California? What's your point?
I specifically addressed your pessimistic comment that dismissed upcoming driverless tech due to your experience testing the current DAS (driver assist systems) built by non-tech companies. I question the relevancy when tiny companies like Comma.ai have demonstrated that widely available deep learning algorithms allow a single guy in his garage to surpass Tesla's own tech. Therefore if that obsolete tech (which is itself is using 3yrs old sensors) is your point of reference plus some 10yr old videos... I'm not sure how relevant that is to what tech we can anticipate in the next 3yrs.
George Hotz' talk opens with the claim that "Tesla's and other companies current DAS tech sucks. They don't even capture the video/sensor data from the rides". Which I fully believe. It shows the wide disconnect between the limited DAS tech being rolled out by car manufacturers and the true capabilities of current technology from people working on real autonomous vehicles. Instead of glorified cruise control with lane switching.
The real question and risks seems to be around price and implementation details. Not whether or not the technology will actually work as intended within the next few yrs.
[1] https://oshkoshdefense.com/technology-1/unmanned-ground-vehi...
For all those "truck driving simulator" games, maybe there would be a market for "drone driver simulator" where you are dropped into random bad situations and have to successfully navigate out of them.
Accuracy also seems like a problem: the GPS often has no idea, e.g., that I'm on an access road versus the main road, or precisely where the next turn is.
I get the feeling we are going to have to add some "beacons" to the current roads to deal with local conditions and just provide a decent verification.
If you are taking a long trip, or hauling freight, then I'd rather get out on the open road, turn on the AutoPilot and take a nap, or do something more productive with my time.
As an example, consider freight from LA to SF. The driver would still drive until he leaves LA, the truck would drive up the I-5 and the driver takes over again south of SF. In the end, the driver only had to drive for 30% of the time and could be more alert during that time (and spend his time with sth useful in the meantime).
I would have to look at actual figures, but my gut feeling given the truck routes around us is that might be suspect. I'll echo galdosdi on the rest, the interstates aren't exactly easy.
This has been well understood for a decade or more.
http://www.curbed.com/2016/8/31/12691516/self-driving-bus-ve...
Video of it in Duabi: https://www.youtube.com/watch?v=qgl0dJ6vRyQ&feature=player_e...
Here's a possible relevant quote:
>I confess that in 1901, I said to my brother Orville that man would not fly for fifty years… . Ever since, I have distrusted myself and avoided all predictions.
— Wilbur Wright, in a speech to the Aero Club of France, 5 November 1908.
Of course, what that doesn't enable is eliminating the need for a licensed, competent human to be present. So it doesn't allow for any of the shared automated car scenarios that get people excited.
I do know a lot of smart people who are convinced that we're on this exponential technology improvement path and nirvana really is only 5 or 10 years away. But I listen to other experts and just look around busy, congested cities and getting computers to the last 5% looks really hard.
If the intended users are anything else than specialists, it's very rare that anything that is not either fully automated or not noticably automated works out. In fact, even if the user is a specalist - such as a pilot - it often doesn't work out very well. People will misunderstand the system, trust it too much, trust it too little, rely too much on it.
The only solution is full automation. Noting else is going to work.
It was mandated because it improves safety. Automatic braking is scheduled to be a standard feature in the next 5-6 years.
https://en.wikipedia.org/wiki/Collision_avoidance_system
I'm happy to concede that these are sort of uninteresting in the context of autonomous vehicles, but they are very much computer systems that assist drivers.
I just didn't realize that I definitely think it's safe to safe to say that there is not going to be any such thing as a computer-assisted human driver in any significant numbers. was not intended to be a categorical statement...
>"Confession time...
I really want to see fully autonomous vehicles but I don't really believe we'll see them on the road in the next 50 years. I think Google's concerns with being able to reliably pass control between the driver and the computer is warranted. I think that the technology will get stuck in the equivalent of the CGI uncanny valley where they find that the technology has to leave more control with the driver than its capable of to keep the driver engaged but never is capable of full control. My guess is that these systems will evolve into assisted driving technologies that will use force feedback to the driver that will suggest the sanest path but won't take full control until the driver is outside the envelope and will predominantly be used to extend the window where baby boomers can drive and also save inattentive and unsafe drivers for themselves. In other words a drunk behind the wheel is still going to look like a drunk behind the wheel but just less likely to kill someone.
I expect the technology will be displace drivers in military convoys but I don't think it will be good enough for general use. Even if the technology is close I don't think the safety will be as good as the "augmented human" model which will also rapidly improve and so insurance and regulation will continue to hamper rollout.
Would love from those in the know to tell me I'm wrong. I want to believe!"
A lot of newer cars have lane assist and other features. I think a half-way there approach works fine. The same way pilots have the option to use auto-land and auto-pilot.
Transmission is completely automated. You never have to handle anything that has to do with it.
ABS and TVS changes how the car functions all the time. You have no control over it whatsoever, and it never hands over control to you.
In other words, none of those contradicts anything I said, and I have no doubt that potentially more things like that will come. However, they will still fall under the same two categories: full automation or non-noticable automation.
But those things don't uniformly work fine, unless you ignore human factors. The big danger here is https://en.wikipedia.org/wiki/Risk_compensation especially with regards to encouraging drivers to feel safe about not paying attention temporarily.
Things like ABS and traction control seem to work great in this light, providing a benefit at little cost -- because (unless they're goofing around on a track or parking lot or something) the driver rarely notices it outside of a rare emergency and thus cannot adjust their behavior to compensate.
Automatic transmissions, lane maintenance, cruise control, "tesla auto pilot mode" etc on the other hand clearly make the car feel easier to drive and harder to crash every moment they're being operated, and make the driving task less demanding of attention, providing plenty of opportunity for the driver to get used to the safety feature and start using it as a crutch that allows the driver to pay less attention. And inattention is the main cause of wrecks.
Even road features themselves often have this effect -- making things seem safer makes drivers actually driver more poorly, and vice versa. See for example http://www.accessmagazine.org/articles/fall-2012/slower-road...
Maybe these features save more lives than they doom, but it's certainly not a given, so as they are introduced they really need to be rigorously analyzed to ensure they have the hoped-for effect.
Except allowing humans control is, by definition, less safe. Google's self-driving cars have had a total of 1 partially at-fault accident after 1.5 million miles of driving. While their accident rate is about the same as the average driver, what isn't the same is that humans caused all but 1 of the 17 accidents the cars have been involved in.
The software is well under 50 years out.
And even for the examples they did show, it would be more interesting to know the rate at which they handle them successfully. So sure, they handled the guy with the reflective vest and a huge ass sign directing traffic that one time...will they do it the next time? Will they handle the case where the construction crews aren't doing everything the way they should?
Furthermore, when autonomous vehicles become common, human drivers will also end up in fewer accidents because they won't have to worry about the unpredictable behaviors of other human drivers (such as suddenly changing lanes without signaling).
Edge cases are important, but at the end of the day, they're just that: edge cases. By definition, they will be very rare.
I'm sure they've got some reasonably safe behavior for when it doesn't know what to do (do nothing!), but that doesn't exactly make for a great transportation device. Staying home is safer than that and just as effective at not getting you anywhere.
And while edge cases are by definition rare, how rare are they? I'd be willing to estimate that when I drove trucks for a living, I'd handle upwards of 100 scenarios per day that were not "by the books". Remember that 99.9% of uptime still means you are having downtime 1.5 minutes a day. If they are really bad at handling the edge cases at the 0.1% level of rarity, that's still enough of a chance to fuck up at least once a day.
Citation needed.
I will not see "self driving cars" not just as replacement of our Camry and Prius but it could be modified to do lot more things we currently have not imagined.
One angle I can think of is that building an environment designed specifically for such vehicles. Currently most of the challenges come from the fact that the car has to run in an extremely unpredictable environment. What if we could build isolated lanes or even exclusive roads for such vehicles ? What if we could build housing complexes, airports, parking lots etc.
[0] Kobe New Transit in 1981 and Lille Metro in 1983
Because it's a kneejerk political assertion of a common bogeyman provided with no evidence whatsoever which fails to explain why there are GOA4 systems in production including in countries with a history of adversarial labor relations?
TL;DR: because it's a worthless pandering comment
They're balanced (or perhaps more than balanced) out by the college-aged libertarian millennials, so that isn't a good explanation.
Driver handoff is an optional feature. Cars could just stop instead. New cars have to have autonomous braking now (or very soon). A timeout on handoff can just have the car slow down to stop (which will trigger cars behind it to stop too). That's a compelling reason to have the driver want to pay attention enough to take the wheel before apologizing to everyone behind them.
I think the typical low information futurist brigade is driving this narrative and they haven't critically looked at what fully autonomous really means. The hype will wear off soon as we keep seeing crashes and other issues. I think the idea that cars will go from dumb to autonmous overnight is foolish. We'll have years/decades of smarter driver assist and maybe someday allow autonomous after a very long process of working out the kinks, certification, regulation, etc. Even then it'll probably only be highway driving or in low congestion areas far from pedestrians, strollers, and cyclists.
There are probably other use cases that'll be approved sooner than later like slow moving garbage trucks with autonomous robot arms picking up trash cans. Or autonomous street cleaning cars. That's assuming the crony public sector unions allow them.
I really want to see fully autonomous vehicles but I don't really believe we'll see them on the road in the next 50 years. I think Google's concerns with being able to reliably pass control between the driver and the computer is warranted. I think that the technology will get stuck in the equivalent of the CGI uncanny valley where they find that the technology has to leave more control with the driver than its capable of to keep the driver engaged but never is capable of full control.
My guess is that these systems will evolve into assisted driving technologies that will use force feedback to the driver that will suggest the sanest path but won't take full control until the driver is outside the envelope and will predominantly be used to extend the window where baby boomers can drive and also save inattentive and unsafe drivers for themselves. In other words a drunk behind the wheel is still going to look like a drunk behind the wheel but just less likely to kill someone.
I expect the technology will be displace drivers in military convoys but I don't think it will be good enough for general use. Even if the technology is close I don't think the safety will be as good as the "augmented human" model which will also rapidly improve and so insurance and regulation will continue to hamper rollout.
Would love from those in the know to tell me I'm wrong. I want to believe!