For some values of 'human' e.g. alert, sober. Which is sadly, a minority?
Its arguable that self-driving cars are better already. What we need to do is, secure the roads so they won't get confused (fence them & remove pedestrians/bikes). I know, we can't do that everywhere but we could do it for long-haul etc. Like freeways already do.
I keep thinking the key is to make 100% of cars self-driving, and networked so they can all communicate and negotiate, which takes the unpredictability and lack of attention of human drivers out of the equation, and then pedestrians and cyclists will adapt. If this was the case, then you could probably get rid of traffic lights and speed limits...but it's the kind of solution to accommodate software solutions that could handle it successfully (I don't like the term AI, because I don't think there's any such thing on a philosophical level). The only problem then remains with the small percentage of drivers left that are determined to drive, like sports car users and motorcyclists, who could interrupt the system by taking advantage of the stability afforded by it. Again, this could be resolved by fitting them with beacons, say, that communicate information about where they are and where they're going, but it'll be a hard sell. I'm pretty sure that most people aren't bothered in the slightest about driving themselves to work every day and operating a steering wheel and pedals to do it, if you could offer them a solution where they could be on their phones and don't have to drive themselves.
It seems to me that self-driving cars can't be just as safe as the status quo, but have to be far far better. It's an unreasonable need, but human nature.
I'll bet that to really make it work well, you also need to redesign roads to suit the new cars.
In any case, I'll believe it when I see widespread/universal adoption of self-driving in constrained environments (mining vehicles, warehouses, storage yards). Step two will be things like garbage trucks (slow, expensive, otherwise automated, phone home when it gets in trouble).
How about autonomous on the highway, turnpikes, etc. And human-operated on streets where a car might ever need to yield to pedestrians or cyclists outside of red-lights.
The self driving car doesn't need to be perfect, it needs to be better than humans. That is a much lower bar. Self driving cars have already demonstrated their ability to handle traffic better than humans, but there are other situations where they are much worse.
The issue is that self driving needs to be better than human drivers by a considerable margin. The public will not tolerate an AI driver that makes the same mistakes as humans, they expect better. To do this it will need access to sensors that humans don't have
How about a simpler solution? Designate certain roads as autonomous only to assure safe operation among other vehicles (for example highways do not allow pedestrians, bicycles, etc; we could require certain roads to be autonomous only as self-driving vehicles become more commonplace), and maps should be updated to identify all possible pedestrian crossings to drive pessimistically?
Sure, but this particular problem seems solvable. A self-driving car can have much better sensors than a human. I.e. maybe we should expect self-driving cars to see in total darkness. That's something we, the public, could decide is a minimum requirement.
It's a bit frustrating because if all calls were mostly autonomous we could likely have less traffic deaths within a few years. Self driving cars are already great at making snap judgements. Better than most humans. They also make a lot less stupid mistakes than humans do.
The edge cases are the hard part. A human knows what to do when a random construction worker or cop is directing traffic. Self driving car? Not so much.
Self-driving cars need to be safer and more reliable for the regulators and mass to adopt. Humans are pretty bad drivers with only visual + audio information when one of both are impaired by natural conditions (weather, night time etc.) or artificial conditions (DUIs etc.).
By introducing more dimensions of sensing (LiDAR, Sonar etc.) as some others put it, it should theoretically help judgement during driving and thus improve the safety aspect.
One solution to bridge the gap between 'mostly' self driving and 'totally' self driving is to increase the safety systems.
Currently, safety systems for the people in the car are very good. But for people outside the car, they are very bad. If you can make it such that the car is very safe for all the people in a possible accident, then maybe the AI problem won't be such a problem.
Granted, that's a hard problem too, but we're pretty good with the theorems and modeling that goes into safety. It's more of an economics problem, not a design one.
Part of self driving is dealing with those human drivers who are terrible drivers though.
This is what makes the whole thing so hard. If we could have dedicated roads for self driving cars it would be easy peasy. No need to interpret road markings, signs, other vehicles and their intentions. Because it would all be communicated digitally over radio. Steering a car is total peanuts for a computer.
It's the unexpected stuff that makes it hard. Like that Uber car that killed that lady that crossed the road with her bike in the dark at a totally unexpected spot. A human is better at those things because they're caused by other humans who aren't logical.
But such matters are really hard to test and evaluate for precisely because they are unexpected.
I remain convinced that "real" self driving (as in: go ahead and sleep in the backseat) will never happen without changes to road infrastructure and possibly some sort of segregation between robot-driven cars and people-driven cars.
Things like traffic signals that actively communicate their status to nearby robot cars (more than just a red lamp that can be occluded by weather, other vehicles, or mud on the camera lens). Or lane markings that are more than just reflective paint, but can be sensed via RF. Rules around temporary construction that dictate the manner of signage and cone placement that the robot cars can understand. The cones might have little transponders in them, I don't know.
But without a massive leap forward in AI capability, our current road system—optimized for human drivers over the past century—is not going to work.
If we can't make the cars just as smart as an alert and capable driver, then maybe we need to meet halfway and make the roads a littler "dumber" (simpler) to accommodate the robots.
If self-driving cars start to catch on, it seems that the standards for roads must improve. There are some roads in my area that could be an excellent Turing test. If you follow what the lane markers are telling you to do, you die and are probably a robot. If you ignore the markers and survive, you are probably human.
I'd rather increase road standards and accelerate the evolution of self-driving technology than require the development of technology that is capable of driving on today's terrible roads. I know this is not a binary choice -- realistically, both would happen. But I don't think anyone's even talking about better roads for self-driving cars. If even just major roads had to be maintained to a minimally machine-readable standard, it would be a good thing for drivers -- whether human or machine.
My impression looking at self-driving cars given these incident is that the whole thing was a pump scheme. They are simply not ready. Roads are not rail tracks. They are highly random and dangerous. You'd need something with a high level of awareness and not just visibility.
Humans have that awareness and lack on visibility (you can't see 360degrees, you can't see a wide range, you can't see beyond your car Caracas...). The awareness was enough to let us move but the lack of visibility did certainly kill people.
The best we could have done today was a system that helps human visually. Then the system reports back how human interacted with that particular situation (ignore, brake, etc...) Then maybe 5 years later of work we could have had something.
So far the focus was primarily on what it would take to produce an autonomous car, which is primarily a technological challenge. But what I see as a harder problem to solve is the one imposed on all humans participating in traffic: we will have to strictly adhere and be very self-disciplined in traffic, in order to make sure that the non-humans get the correct message.
Imagine I am a cyclist and I don't signal but I swiftly swerve to the right. Unless the self-driving car is already at a slow speed, it would impose an undue risk.
Or, it is hard to expect that humans will rigorously adhere to this traffic discipline.
This is why these cars will drive slow for the foreseeable future and will frustrate a good chunk of the drivers.
Unless it gets dedicated lanes, the way bicycles do.
I appreciate your enthusiasm for autonomous vehicles, but to be fair, there are many things that could be done to improve road safety without self-driving tech. Personally I love the independence and self-determination that comes with driving. I'd rather roads were much safer. I don't really want to be part of the WALL•E-esque future you describe. (But I'd happy to hear a counter-argument or further elaboration of what you'd like to see happen.)
Driving on roads with human drivers, bikers, pedestrians, animals, freak weather events, maintenance and construction workers, and all manner of situations is - in reality - a problem of unbounded complexity. Unless we create infrastructure that radically reduces the complexity, like invisible “tracks” and banning as many variables as possible (other humans) then nothing short of AGI will get us close to the current below-average human driver.
Self driving cars is Silicon Valley hubris and hype at its finest. I genuinely feel proud about the computer scientists who have wrangled millions of dollars from investors, truly excellent work, I wish all engineers could get that kind of money. Hopefully the billions keep flowing.
I think the problem with that is that to be generally useful, autonomous vehicles need to operate in environments with plenty of entities that are not participating consistently in the standard:
Human driven cars, pedestrians, bicyclists (about half of whom dart around unpredictably, ignoring other vehicles and traffic control signals), motorcycles (like bikes but faster), the weather, etc.
So, yes, much cheaper and easier to implement, but also not very useful because they won't be able to drive very many places people want to go.
Its arguable that self-driving cars are better already. What we need to do is, secure the roads so they won't get confused (fence them & remove pedestrians/bikes). I know, we can't do that everywhere but we could do it for long-haul etc. Like freeways already do.
reply