Seeing Issues with Self-Driving Cars


It’s been a long time since I’ve filmed
outside… or in my car. Though, according to some people I won’t
have to be sitting here for long. Which reminds me, I recently picked up a
large British audience because of a few videos I made, so let me uh… This should make them feel a little
more at home. And is oddly appropriate since this is what
many people think this is what the driver’s seat of a car might look like soon. No steering wheel, maybe just a HUD… it’s
pronounced HUD, Heads Up Display, not “hood.” And I won’t have to do anything more than
just sit here and sleep or… sing. Really, we don’t have an embarrassing shot
of me singing? But there are many problems and challenges
that self-driving cars must overcome first. When artificial intelligence, first started,
the designers thought to themselves: “Hmm, what’s the most difficult thing we can teach
a computer to do?” What answer did they come up with? Chess… and checkers and go, board games. But not just any board games, these are games
for super nerds, so if they can beat super nerds, they must be really smart. Turns out, while these games may be hard for
you, they’re really easy for a computer. Like laughably easy. Deep Blue beat the beat the best human chess
master in 1997. That was well before most of you had internet,
if you were lucky enough to have internet, it was still dialup and you hoarded those AOL CDs
with thousands of hours. The game checkers was declared officially
dead in 2007, because a computer had memorized every possible game situation, all 5 times
10 to the 20 positions. In 2015, Chess was likewise solved. And in 2016, a computer beat the second best international Go player… Board games are easy. You input the rules, maybe let it watch a
few thousand games, and that’s it. It doesn’t require any motor functions or
sensory input. So what people didn’t see coming was how
difficult it is to get a computer to see. Seeing requires much more than just photon
receptors. That’s why you have an entire lobe, roughly
20% of your brain, dedicated to it. Colors are relatively simple and not all that
difficult for a computer to figure out… so they won’t be the focus of this video
– but I’m… building to it. Shapes are also relatively easy but edge detection
is not. So when a computer sees an image like this,
it sees a bunch of colors and shapes. But it has no idea what any of it means and
can’t tell the objects from the background. Is this an object? Is this an object? Is this all one object with multiple colors? But you can see it, easily. And without having to think about it. Because, one of the primary purposes of Area
V1, also known as BA17, is edge detection. It’s one of the first things that the brain
does with visual information. This problem has been mostly figured out,
but I’ll get to how in a bit. A problem that will never be solved though,
is how to teach a computer to see depth. It has no idea what is closer. This… or this. Again, you can, with little to no effort. When you see optical illusions like this,
you know something isn’t right, but a computer has no idea. First, let’s go through the monocular cues,
that is one eye, for depth perception. When you look at this picture, you see the
horizon. Your brain uses this as a major clue. The closer it is to the horizon, the further
away it usually is. And the higher up or down, the closer to you
it is. We also look for parallel lines which converge
the further away they are. Also the further away you look, the more atmospheric
haze there is, which makes things have less contrast and appear more blue than they should. But a more obvious monocular cue is relative
size, which is how big or small something is compared to something else that should
be similar… like in this illusion. They’re all the same size by the way, but
because of their positions they appear hilariously misproportioned. Or, my hands, one of them is bigger than the
other – not because I’m a freak, but because one is closer to the camera. Another obvious cue is occlusion, since my
right hand is obscuring the view of my left hand, it must be closer. Which was also the case with the cars. But that’s just how you do it with one eye. You don’t actually see this. Because you have two eyes, giving you binocular
vision… so you see this. The focus of your two eyes gives you depth
perception. Because of a fun little bit of math that your
eyes and brain calculate without your conscious awareness. The angle of your eyes is calculated when
you’re looking here… and when you’re looking here… and your brain essentially
does triangulation to figure out that this point is further away than this point. Granted, you don’t actually perceive this
blurry mess unless you take off your glasses during a 3D movie, but in real life your brain
cleans up the image a lot. While you may think that you see in 3D, you
actually only see in 2D and your brain creates a 3D space using that 2D information. And we’re not even going to get into the
motion cues because, we’ve already complicated things enough… and… I mean c’mon. Maybe that’ll be another video. All of these are things your brain does without
even thinking about it. Many of the monocular cues can be programmed
into a computer, but not all of them. And just like how your brain can be fooled
by various illusions, so can a computer, which is why a computer cannot rely entirely on
vision to determine depth or distance. It needs some sort of outside help. Many of you might have jumped to the idea
of GPS or satellite imagery. Unfortunately, GPS doesn’t provide you with
any images, and it doesn’t tell you where you are. It tells you where IT is, and your computer
makes its best guess on your location, which is usually only accurate within a few feet. Which doesn’t work when you’re talking
about piloting a one ton hunk of metal at 60 miles an hour. Satellite imagery also doesn’t work because
it isn’t live, which is pretty much what you need when navigating a busy street. And even if you somehow manage to fix that
problem you’ll never overcome weather, buildings, and trees and stuff. So a driverless car needs to use localized
vision, along with something else to perceive distance local to the car. Which we’ve actually had much longer than
we’ve had satellites. Echolocation, such as sonar or radar. Basically, it sends out a signal and based
on how long it takes for the signal to reflect off the object and return, it can figure out
how far away it is. The problem, is that this information is limited
and not quite how you see it in the movies. For an image like this, radar will return
an image like this – just like with visual information, it’s only two dimensional,
and will only tell you how far something is from transceiver at the level of the transceiver. So if it’s on the roof of your car, that’s
not very useful for looking for how far away another car’s bumper is. So have one on the bumper, obviously. That’s on top of the fact that it only tells
you how far away something is in that moment, and not what direction or how fast it’s
travelling. Which is made even more complicated when you
add in the fact that you are likewise moving. The technology to do that does exist and has been in use for sea and air travel for decades. But those are long range with far fewer vehicles
and little environmental obstruction. So okay, all of these are challenges that
have been or can be solved. So let’s look at some issues with vision
that have not yet been solved. What is this? Right, it’s a bicycle. What is this? It’s still a bicycle, c’mon. Okay, what is this? It’s still a bicycle, this isn’t rocket
science, it’s pretty easy. Yeah, for you. For a computer, it’s infinitely difficult. Object recognition is by far the hardest thing
to get a computer to do. Especially when you consider that 3D objects
look completely different in a 2D image from different angles, even in perfect conditions. You may remember this video from a few years
ago of a robot navigating around the world and interacting with objects. You probably mostly remember it because of
the jerk with the hockey stick. It’s all pretty impressive… as long as
you aren’t aware of the difficulties of getting a computer to see. You know what a cognitive psychologist sees
when they watch this video? A robot interacting with a bunch of QR codes. Those QR codes tell the robot what the object
is, it’s distance, and it’s orientation. Whether it’s a box or a door, if they want
the robot to interact with it, there are QR codes stamped all over it. Sorry for ruining the magic. What are you talking about? Computers are way good at recognizing things,
like how facebook recognizes me in all those pictures or all those snapchat filters. Yeah, I mean they’re good at faces, I’ll
give you that. But faces are pretty easy, there’s a pretty
standard pattern for those… your brain is wired to automatically think this is a face
– it’s not… it’s a chair And it’s not really like recognizing faces
is all that useful for driverless cars. Computers are getting better at recognizing
objects though, and you’re helping. Whether you’re playing Google Quickdraw,
or doing one of those annoying new Captcha images where you select all of the squares with a stop sign, you’re teaching computers how to see. You are helping our future machine overlords
recognize objects. However, even with all of this learning, it
still has to match what it sees with a previously learned template. It’s not, like you, able to figure out what
new objects are and what they mean on the fly. For example, say you come across this sign. It might only take you a second or two to
skim it over and realize that it doesn’t apply to you and continue driving. But a computer? The first thing it will recognize is the shape
– it sure looks like a stop sign. It’s red and white, just like a stop sign…
it has a little more to it than a stop sign but, just to be sure, I better stop… in
the middle of this street. That could cause an issue. Or perhaps this situation. Clearly, that’s a stop sign, but it’s
covered with a trash bag. Whatever internet traffic uplink it’s using
says there should be a stop sign here and there it is. But it… it’s partially covered with a
trash bag. Again, you can look at this situation and
quickly figure it out. You’re supposed to follow the temporary
green light and ignore the stop sign. But a computer, especially one that’s never
encountered this before, won’t be able to react as easily. Say the temporary light wasn’t there. Is a trash bag and some duct tape all it takes
to fool a self-driving car into ignoring the sign and just flying through? If so, you’re going to see a lot more teenage
pranks and youtube videos like this show up. And what about that detour sign? Your GPS says you should go straight. You might know that you can safely go straight,
but a computer sees the sign saying it must go right. Maybe there’s an obstruction ahead? These are all things that you can quickly
figure out. But a computer has to obey a set of rules
and when presented with something outside of its ruleset, it may not know how to react. Which is why I’m not very impressed when
people bring up Google’s self driving car that drove up and down the highway on its
own a few years ago. Driving on a highway is easy, all you have
to do is stay between the lines. There are very few dynamic situations, very
few unique situations, and relatively few challenges. It’s so easy a monkey could do it… actually,
it’s so easy a dog could do it. This isn’t a prank, this isn’t a joke,
this is an actual dog driving a car. There’s an entire channel dedicated to showing
various dogs learning how to drive cars, it’s not that hard. Obviously they’re on a track by themselves,
heavily supervised, but they are staying in the lines. It’s not that hard, and it’s not that
impressive. Highway driving is so easy you regularly completely
zone out and stop paying attention, and most of the time everything turns out okay. Not like in the city where you’re on constant
guard and see dynamic, unique situations all the time. Which brings us to the last major hurdle that
driverless cars must face. Once you’re able to get it to see properly
and understand what it sees, you then have to tell it what to do with that information. Let’s say you’re driving along and you
come across this situation. Again, you, as a human, can quickly figure
out the context of this situation, and probably wouldn’t stop. A computer, on the other hand, wouldn’t
know that this person was just about to turn right and get into the driver’s side door. According to this person’s current trajectory,
if the car doesn’t stop now, it’s gonna hit them. Does it assume that this person is fully aware
and acting safely or does it stop, possibly causing an accident with the car behind them? That’s a simple situation, a very simple
situation. Let’s say that the car is driving along
on a two lane road and realizes that it’s brakes are out. I don’t know, maybe a line got severed or
a wire shorted out, it doesn’t matter – it’s rare, but it’s not unheard of. Coming towards the car are two motorcyclists
who are dangerously riding side by side in both lanes. Your car must now choose who to hit. You, as a human, can freeze up, yell “Jesus
take the wheel!” and let physics decide who lives and who dies. A computer on the other hand, can’t. Not making a decision is a decision to do
nothing, which means that the car will hit one or both of them… which means that the
car decided to hit one or both of them. There is no scenario where the computer can
claim to have been so flustered that it couldn’t make a decision. It could decide to follow the law and strike
the person who was travelling in the incorrect lane, that’s one way to do it. Or, we can make the situation even more interesting
by pointing out that one rider is wearing a helmet and all their protective clothing,
while the other is simply wearing a t-shirt and shorts. Your car may decide that this person is more
likely to survive a collision – although very slimly more likely – and therefore
steer the car into that person. Which would paradoxically make it less safe
to be wearing a helmet. Or, we could point out this tree to the side. It could avoid hitting both riders and instead
elect to crash itself into the tree… just injuring you. You’ll likely survive while the riders probably
wouldn’t. But who is the car supposed to protect? You? The owner and operator? Or some bonehead Harley rider who wasn’t
obeying the law? Some might say that as the owner, the car’s
main directive should be to protect you and the passengers. While others might say it should protect as
many lives as possible. But given the choice, if there were cars on
the market that safeguarded all life and cars that just protected you and your passengers…
you might be more inclined to buy the one that places you and your family above others. Let’s pose another situation. Say you’re at an intersection, and your
car wants to make a right turn… but there’s a line of school children currently crossing
the street, all holding hands, single file. So you’re patiently waiting. But another car coming down the road, has
hit a patch of ice or has its brakes and steering go out, whatever, it doesn’t matter, the
point is that the car isn’t stopping and no longer has control. It’s also a self driving car, and using
magic, is alerting all other cars in the area about its situation. If your car is designed to only protect you,
it’ll probably sit tight… and force you to watch something so horrifying you’ll
never see the end of the therapy bills. If your car is designed to protect as many
lives as possible, it might pull forward into the intersection… stopping the car from
plowing through all those kids… but you’ll be t-boned and your possibility of walking
away from this accident is pretty low. These are the situations that driverless cars
will be forced to have to make decisions on, and they are incredibly tough decisions. Not to mention the fact that I’ve only given
you a small handful of the literal infinite amount of possible situations. I certainly don’t want to be the one writing
the ethical and moral codes for self driving cars… but someone has to… especially if
we ever want intersections to look like this. Where there are no traffic lights, all the
cars are driverless and are simply communicating with each other with hyper efficiency. And it’s absolutely impossible. First of all, it requires that every single
car on the road be self-driving. If there’s even one manually driven car,
game over. Which then also means that your car must be
self-driving all the time. If you have the ability to switch it on and
off, an intersection like that will never work. Which means that old man river out on his
dirt road would have to be using a self driving car. We can get around this situation by saying
that only on certain roads, auto pilot must be enabled, fine. But let’s say you’re on one of these auto
pilot only roads, and you’re late for work… when this happens. Your HUD tells you that an emergency is occurring
on the road so all travel is currently halted. Nevermind how furious you’ll be over the
fact that the government can just seize control of your car… you’re late, so you flip
the manual override and decide to proceed anyway… and congratulations, you just caused
a collision. Don’t act like this situation is impossible,
how many people do you know who have driven through a closed highway because “the weather
was bad.” If people can break the law in order to save
themselves time, they will. But let’s go back to this intersection and
assume that all cars will always be self driving, with no possible manual override. This intersection is still a disaster waiting
to happen. Let’s completely set aside the idea that
anyone would ever go to this intersection with malicious intent, even though those people
have always and will always exist. And we’ll assume that all of these cars
are completely unhackable, again… we’re assuming perfect conditions. Imagine a tree branch falls in this intersection. Or a tie blows out. Or a truck’s unsecure cargo falls out. You’re looking at a several car pile up…
even with AI that can respond instantly. Also, having traffic flow like this renders
the intersection completely useless to pedestrians and bicyclists. There’s an easy solution of course, a four-way
foot bridge. Which likewise dramatically increases the
likelihood of something or someone falling into the intersection, accidentally or not. But again, in order to achieve that perfect
flow of traffic, everyone needs a driverless car. Cars aren’t like phones, where people get
a new one every year or two. Cars last a long time – like 15 to 20 years. Most people don’t get a new one until their
current one is broken beyond repair. So even if by some miracle, all of the technological
and ethical hurdles are overcome in ten years – which is extremely generous, they totally
won’t be – and they stop selling manually driven cars that same day… without government
intervention, it would still take another 15 to 20 years to phase out all of the manually
driven cars, excluding the antiques of course, because you’re never going to take those
away from people. On the topic of government intervention, we
also have many legal issues to work out with driverless cars. Just to throw a few out there. Who is at fault when a car decides to hit
someone? When you’re the only person riding in a
self-driving car, are you allowed to be on your phone? Sleeping? Drunk? If you’re required to be awake and attentive
the entire time, doesn’t that kind of ruin the point of it being self-driving? Self driving cars will happen, don’t get
me wrong. They are coming. But if you think that they’ll take over
the roads in the next ten, twenty, or even thirty years, hopefully now, you know better.

, , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , ,

Post navigation

100 thoughts on “Seeing Issues with Self-Driving Cars

  1. Would you want a car that protected just you – or as many lives as possible? Who do you think the car should hit? What about the legal issues like sleeping and drinking while riding?

  2. Chess hasn't been solved like checkers? There are endgame tables for up to a certain amount of pieces, but it's not a solved game, i.e. that there is certain outcome given perfect play

  3. I don't think they will take over soon. But I think they will be street legal pretty soon I think, maybe 2021-22. The tech to make them completely safer than human drivers in nearly all circumstances probably by 2028 at which point shipping companies will push for unoccupied shipping vehicles. A few years after that Taxis and Busses too will push for driverless options. At that point there will be no legal reason you couldn't sleep in your car. In fact they will be so much safer than human drivers you may begin to see a loby to remove human overide altogther. AI's will still make poor choises and have accidents, but if they do so 50 times fewer than humans we won't have much of a case to stop them.

  4. Or just have everything mapped out, insert it into the car's computer, and the cars can communicate with eachother

  5. As a Brit, I felt much more comfortable when you sat on the right side of the car… It's the best side

  6. We need to gradually transition into fully self driving cars, by having cars that are partially self driving and allow drivers to overrule any decision they're about to make. They should have manually driven cars with self driving capabilities, that tell you what they see and what they would do, without actually doing anything. Researchers should collect this data for years and see how accurate it is and where the biggest issues are.

  7. Haven't you seen Knight Rider? Computer will simply engage turbo boost, and jump over the obstacle, engage ski-mode and go between the obstacle. Or just disable it with micro-jam. If it's KARR, you can just shoot lasers at it.

  8. 11:14
    Program it to shut off the motor and shift into 1st.
    If it's an electric car, maximize regenerative braking until it can safely reverse motor rotation.
    Or just hit both guys. Why not?

  9. ok i watched those dog driving videos, to say they taught them to drive is a stretch
    but you could probably teach a dog how to do more than take a right turn. like maybe use the brake…

  10. Chess has NOT been solved. I don't know where the hell you got that from. It's been a long time since a human beat the best computer at even a single game, but even that is a far cry from solving the game.

  11. Honestly the situation with the intersection is pretty easy to smooth out. You just have to give up that hyper efficiency and settle for mere inhuman efficiency. You could even have it mandatory for all cars to maintain double safe stopping distance and still have clearer, faster moving traffic then humans could ever achieve.
    As for miss reading signs, well, you sort of answered this question earlier: cheat. If a computer finds it easier to recognise QR codes over symbolic signs then use those. Then it would be a relatively east case of slapping a QR code over the current one to re direct the car. Using a special coding system, possibly related to location would also make it so that these QU codes are unique to that place thus making it a lot harder to create "prank" versions, not to mention having the car as well versed as it can be at double checking conventional signs and road markings not to mention reading the environment to see if things match up properly to avoid such circumstances.
    To help self driving cars contend with conventional cars once they become the majority you could make it law that all manual cars have to have transceivers that tell them about that car and what it's doing at each given moment. Not as good but enough for it to easily compensate with a little more cautious driving and if there's a vehicle completely without one it could be taught to recognise those cars and be extra careful.
    As for the ethical dilemmas you've brought up… well, that's probably going to have to be put up to lawmakers to decide pre-emptively. Maybe they will decide that self driving cars should prioritise the safety of the greater number of people over the few passengers in which case self driving cars might not be so popular from then on, well, so be it.

    I mean, what I've said hardly solves every problem you've brought up and doesn't entirely solve the ones you have suggested but I've only been thinking about this for a hand full of minuets. There are solutions out there. Honestly self driving cars is a a solution to the ludicrous number of road accidents (usually a majority most ways you measure the statistics) caused by human error. Personally I would rather improve the human but hey, cybernetic and biological augmentation is a lot further off and if self driving cars makes it a little safer till we can even if it's not perfect? I think I would rather take that.
    This also encourages the bigger issue of "but sometimes". For example, LED traffic signals are superior to conventional traffic lights it pretty much every way. They use less power, LEDs rarely fail and if they do such signals are composed of maybe hundreds of LEDs so one failing doesn't matter… but sometimes, it snows and the heat given off by incandescent bulbs melts the snow meaning they remain working. Doesn't matter that you can easily integrate a cheep heating element into a LED traffic signal that still uses less energy and only when it's below a certain temperature, the fact it has happened once has permanently tarnished a superior technology's reputation.

  12. Um what the hell are u talking about chess was not ever "solved" if it were we would see the top computers either only white wins or draw every time. But alas leela and other neural network chess programs show us that even in computer chess human like playing and sacrifices and positional controp matter more than the flawed computer materialism.

  13. Wait… is that why Captcha has been asking me to identify all of the major economic centers on a map of New Jersey before I stream my porn?

  14. Some of these issues have been solved for decades… like an INU/IRU/INS/IRS… whatever you want to call it that day.
    I've worked on a vehicle that can lase a target 1000m away, 360 coverage… and determine where it is within an inch. It starts with knowing — to ridiculous levels of accuracy and precision — exactly where the vehicle is, or rather… the laser system.
    Aircraft do the same thing.
    Mary it up to GPS and you have – literally – pinpoint accuracy for your location in the universe.
    I think L4 will have to rely heavily on that, known road mapping, active RF/laser devices to dynamically model the immediate 3D world, and finally some cameras to fill the gaps. But the cameras are easily the weakest element imo… a lot of fuzzy logic going on there. If your radar or lidar says something is a solid object, you can take that to the bank. MTI, phase, doppler, I/Q video, IISLS, TCAS, etc. are also all old established tech through the ATC Radar scene. Bolting TCAS onto cars would do quite a bit. Replace el data with vector data.
    Tesla has it exactly backwards.

    I doubt L5 will ever exist.

  15. Thank you for your views, very informative.
    I've had many thoughts about this Technology, and beliefs on why it just not posible to achieve any time soon.
    You touched on some of the same issues I perceived it would have to face, but expanded to many more which I hadn't even thought of.
    Your knowledge is great, I only wonder how did you get there in the short time you did, because you really look fairly young.
    But thanks again, and I look forward in seeing more videos with your knowledge.

  16. Hot take: The future shouldn't have cars. Super connected public transport is safer and healthier for the environment.

  17. the solution: keep the driver in control via manual override. and no, it doesn't beat the point if you have to be attentive at all times, because it would be perfect for people who CAN'T be (ADHD guys for example), since they would only have to rely on the car for short periods of time, when they space out

  18. another simple thing – big puddles. I'll drive around them. A self-driver will just plow straight into them. Then you got rough roads/potholes. The self driver is not going to try to avoid them……..there's simply WAYYY too many pitfalls. By the way, can you imagine heading straight into intersections with cars coming at you from all direction like in that simulation??!!! I'd have a heart-attack the first day!

  19. Well sucks for you, driverless car enthusiasts, because I am staying next to manual cars, I am not SO lazy, and will drive cars the “old fashioned way”

  20. How people use cars will be different, think of personal metro. also, insurance will take care of ethics. I think its near 5-10, not 20+years.

  21. Have you ever had anything go wrong with your car? Maybe a blinker isn't working, or a windshield wiper, or a brake pad. You could get by if something like this happened…at least for awhile…

    Have you ever had faulty information from your GPS? It happens quite frequently.

    Driverless cars depend on enormously expensive sensors. If even one camera, or one depth perception device, or one piece of GPS information goes bad, your driverless car cannot be driven safely. But hey, you're late for work and the boss is on your case…

    I don't see driverless cars ever happening.

  22. I've got a question. What do self-driving cars do about winter? I live in a city where we have snow cover all winter long. Once the first snow stays in November, we've got snow on the ground until March/April. Very frequently during winter driving, the lanes are basically indistinguishable. The only reason I know what I'm doing is through years of experience driving in this city, and human intuition. Have the computer people figured out how to deal with winter?

  23. We might not have fully automated cars soon but we can make them help us be better drivers – for example – not let us tailgate and rear end the car in front of us or let us change lanes and hit a car next to us etc. These are now being put in many new cars.

  24. Oddly enough computers are slow at processing constantly changing information and unforeseen scenarios as the US Navy found out. On aircraft carriers they used nuts and bolts of differing sizes to represent fuel and weapons loads on aircraft. But they decided to update and use computers instead. The only problem was that the computers couldn't process the constantly changing information fast enough and the computerised system crashed. So they went back to nuts and bolts. An example of if the system ain't broke don't fix it.

  25. The bowling strike dound at 13:25 was so unexpectet hilarious i couldnt resist to stop the vid and laugth for a minute

  26. These are all very interesting points and discussions, but some people look at these problems and use them as actual arguments against self-driving vehicles as a whole. These are problems that need to be solved, not reasons why we shouldn’t even try.

  27. I’ll make my own self driving car so I can turn off A.I mode and do things myself.
    But then I’d be a problem.
    So what’s the point of a self driving car in the first place?!

  28. Hey man. True fan here. Love your videos. But need to sort of hip you to some possibilities that people don't consider:

    1. Early iterations of self-driving cars have to work basically like people do. Future iterations may not.

    2. For instance, self-driving cars can be networked and be required to have transponder capability. (Much like airplanes). Your car won't have to SEE other cars to avoid hitting them. It will just know where all cars are, nearby, how fast they're moving and where.

    3. Visual navigation is fine and self-driving cars can do it. In the future, though, roads can be constructed with inexpensive optical and magnetic and other types of sensors that let cars navigate even better. For instance, your car could be reading mile marker bar codes every 1/10th of a mile. Combined with their data of information (think of wayze) they can know that a traffic slowdown is going to occur in 1/4 of a mile or whatnot. Your car will be aware of the environment to a degree that's impossible for a human.

    4. Visual recognition of NON CAR objects? Man, we're not even in the 1st inning of a 9 inning game, there. As I tell my students "once a computer can do something, even badly, the game is over, because the computer will get relentlessly better, 24×7, forever." Don't judge future capabiltiies by today's technology. Our technology will improve. Insanely. They will exceed human capability. That's not in some distant future. That's in the next 5 years.

    5. Keep in mind that we kill, what, 40,000 people a year on our highways? Something like that. And how many of those are due to drunk driving? Inattentive driving? Sleep-deprived driving? Self driving cars won't suffer from any of that. Once this technology is fully fleshed out, we could literally reduce our fatalities to something like a few hundred people a year. Maybe a few dozen. That's basically like curing a form of cancer.

    So, yeah, don't think of cars in terms of how well they imitate people. A wrench doesn't work because it's exactly like a person's fingers. And technology works because it does things humans can't.

  29. Gosh, I am a fan, but this one did not age well at all. AI is just moving so quickly that pretty much all your points are out of date. This is the danger of playing the Luddite.

  30. Hope that these never become a reality. Not that people drive all that responsibly, but it is possible to tighten the laws about who gets to keep a license. If someone wants not to drive, then he can take a taxi or a bus or a small pickup service that is popular today. Driverless cars are a bad idea in open traffic. They could only work in very limited situations with sever controls in place.

  31. I always hated the idea of self driving cars I love driving and own a 92 Chevy c 1500 and a 97 k 1500 and love to drive them both plus I’m a bit of a gear head that loves shit like donuts burnouts and off-roading

  32. It's highly likely self driving vehicles will be commonplace on the highways in 10 to 15 years. But I do agree that driving within city limits will remain mostly / all human for foreseeable future until/unless a breakthrough in AI and/or quantum computing occurs (which would drastically improve computing power)

  33. Excellent video and points. The biggest issues will be liability and policy and not technology. Boeing is going to pay out $100M for the 737 design flaw. Once self driving cars are on the road there will be accidents and the lawyers will swarm in and take down all the corporations involved. Yes, I am shorting Tesla.

  34. 12:01 here’s I see 1 go in the middle of they lane second go out of the highway and come back in third go of the highway until it stops

  35. i enjoy the video and think you made reasonable assessments, but soon it will be only ethical to have self driving cars. Also…. just because an intersection doesnt have traffic lights.. doesnt mean they cant have cross walks and a button that makes cars pause to allow a pedestrian to cross…. so the whole.. build bridges for every intersection, seems a bit much.

  36. Obviously the car should choose to hit Celty. She doesn't have a head and she's alive. She will be fine.

  37. Ummm.. why don't scientists build a giant sling that launches humans to their destination ? ummm.. where there is a .. giant .. uhhh… ummm.. baseball ?… glove ? to intercept the flying human and land him safely.. this isn't a good idea, like.. at all..

  38. I think this will be something that just gets done within the next decade and THEN we start just fumbling around figuring out all the legal, moral and ethical issues. Tech tends to develop exponentially and not linearily so this will catch a lot of people by surprise when the automation wave hits in full force.

  39. As I recall, sitting back and "letting the car do the driving" ended with a car owner under an 18-wheeler, considerably shorter than before letting his Tesla do the job. A (relatively) cheap human driver would have led to his "continued existence."

  40. Those aren't QR codes! They're fiducial markers- I think they're Aruco, but they could be any of a number of markers in that class. Although they make recognition and pose estimation of tracked objects easy and precise, they're not necessary for this task. They could also use a feature extraction algorithm to recognize the door or other interactive objects and estimate their poses using only the features already in the environment. However, this takes more work than printing out a sheet of paper.

  41. Millions of people who think they can survive a car wreck will not survive the end of this video.

  42. Self driving cars come with a lot of issues, but they’re still almost definitely a lot safer than the bloodbath that is millions of flawed monkeys in cars, even with our current technology.

  43. If you get t-boned by the other car, it could push you into the children. Since your car is longer than theirs is wide, more people could die because of you trying to stop it.

  44. Normally you're pretty close to the base, I don't always agree with you but you're close. This has bad thought process all around, it was obviously a poor subject for you to take on. While correct about getting the vision aspects cleared up (see what I did there!), everything else was short sided and/or a complete misunderstanding of what the real world driving applications do/will entail. I didn't mean to write a dissertation but you had a lot of points that were way off base.

    Much like traffic lights, the government will have some sort information producing devices. Meaning they may have ground based LIDAR placed at intersections and roadways, then broadcast that information to the vehicles in a specified vicinity (maybe not LIDAR but.) That and the idea of multi directional communicating vehicles, all intersections would be totally safe for all pedestrians following the law, without needing traffic lights.

    Despite what a large percent of the population believes, the pedestrian rarely has the right of way. Self driving or not, the pedestrian is the one at fault if they disobeyed the law. A self driven car will be upheld to more strict standards than human based driving, they will also be far better than we are at it.

    The 2 cyclists won't be a dilemma, one of them is breaking the law. Play stupid games, win stupid prizes (same as pedestrian strikes.) A human would have to make these same decisions, and we are not inherently better at it than a computer would be. A human would hit the brakes and hope for the best, the self driving car would too.

    The idiot walking to their car without waiting for the other car to drive by is an idiot. This is not, and should never be viewed as, a situation that anyone should not hit the brakes. Because people are stupid and may just walk all the way across the street because "pedestrian has the right of way" stupidity. The self driven car should too.

    Your children 4 way stop roadway isn't an issue either, because a person driving would not block the intersection either. So what transpires would transpire the same way, person or self driven. The self driven car would see the children way ahead of a person driven car, and because of automobile to automobile communication, the self driven car would probably already know the ice would cause a problem. In the even that it did not, it would be traveling at a way lower speed than most person driven cars at that point.

    Driven cars will cause nearly all the accidents, if not all. When the 65-75% tipping point of self driven to person driven comes, the cries for laws against person driven cars will start. By 85%, person driven cars will become illegal. My thoughts on timeline are WAY faster than yours, because I see ALL new cars being sold as self driven in 20-25 years. I see the 30-35 years being that 85% person to self driven.

    This is long enough as it is but my thought process is that the traffic fines will sharply increase for person driven cars , because self driven cars will break so few laws, and that will get people faster into self driven cars (a lot of cities use traffic violations as a huge source of income and as the law breakers become limited, they will punish the ones still there more.) Plus, people want a chauffeur, so people with enough money to buy new cars will opt for new chauffeur driven ones (more often than not.)

    Most of your points seem to revolve around who is at fault, who gets punished. This is the one real question that would have to be answered, and would come up so few times that it may not even matter. Fundamentally the owner of the vehicle would be the one okaying the trip, it might be a case of making the law that simple. Maybe not totally fair, it would be mostly fair.

  45. The ending of the video is what I hope we get to. Cars using dynamic street laws by communicating with each other.

    This would take awhile to achieve obviously. It does seem weird if people are wasting time trying to teach computers how to recognize something as humanized as a stop sign though

  46. Pulling forward into the intersection could result in more of those kids being killed than if just the out-of-control car went through the crosswalk because now you have two cars flying into the crosswalk.
    Just a couple days ago there was horrible video of a guy on a bicycle being killed by a vehicle that had been bumped out of the way by a speeding, red-light-running moron.

  47. A week ago or so there was an accident in Poland, when one famous journalist drove his car over 300 kilometers, on a freeway… while being completely, utterly wasted. Turned out the reason he managed to drive so far, being that drunk, without any trouble, was that his car did most of the driving for him. It adjusted the right speed and remained stable in the lane. However, he reachead a point where there were some road works, and the signs and road lines were replaced with a temporary ones, organizing the traffic for the time of the construction. The car apparently couldn't tell the difference and got confused which lines to follow, and started to drive through traffic cones. As a result, the guy got caught.

  48. full A.I has always been around the corner since the late 50s. Don't hold your breath unless there is a huge breakthrough in hardware.

  49. Weirdly enough, I DO see in that split vision from ~4:00. Apparently most people don't have that. I guess my brain doesn't fully filter it out, since it only is focused at some specific distance.

  50. Good I really don't want them to come. Getting rid of the ability to drive rids us of a certain independence that is just to satisfying to give up.

  51. Even your camera knows the distance of your hand. That's why it can focus on both of them. This video is full of BS. Sorry.

Leave a Reply

Your email address will not be published. Required fields are marked *