Google's driverless car is brilliant but so boring
bbc.com
bbc.com
I dislike this being described as smarter. So far, what we've seen as described is behavior that is more predictable and consistent. In the scenario described by the OP, in which the car slammed its brakes because of a jogger on the opposite side of the road...that doesn't strike me as a decision based off of what we'd traditionally consider "intelligence", but a strict adherence to a set of rules that could use a little more refinement and complexity.
Even if self-driving cars do dominate future roads, it's going to be an interesting few years of transition...human drivers are going to quickly realize that their automated counterparts are almost always going to follow the rules, allowing humans to cut them off without worry...Maybe at some point, these incidents are going to be logged as evidence (using dashboard cams that are almost certainly going to be standard on all driverless-cars) to prosecute those overly aggressive human drivers?
It's easier for a machine than a human, but it can be learned. I doubt it measurably affects my commute times.
So the machine isn't being "wise" when it avoids road rage, it's just being a machine.
Humans seem to need some kind of training let go of our natural instinct to punish "cheaters". A machine can just do it, as can a stone for that matter. So: artificial wisdom.
I don't think driverless cars will act as individual units for a very long time when the opportunity to have them operate collectively can produce meaningful efficiency gains. For example, drafting on freeways to reduce drag, or being in communication with local streetlights to shape traffic patterns. That's one of the great benefits of moving away from human controlled cars - collective intelligence is possible.
In a way, the rules are a kind of "codified intelligence", and a computer that follows them can be said to be smarter, at least if we defined that based on the results, eg. "able to take a better path to achieve the same goal" (in that case, hurting less people).
Even if self-driving cars do dominate future roads, it's going to be an interesting few years of transition...human drivers are going to quickly realize that their automated counterparts are almost always going to follow the rules, allowing humans to cut them off without worry...Maybe at some point, these incidents are going to be logged as evidence (using dashboard cams that are almost certainly going to be standard on all driverless-cars) to prosecute those overly aggressive human drivers?
Maybe, but there's also the 'glasshole' effect - others drivers will already be anxious about driveless cars, you don't want to add reasons to make them angry at them as well.
You were saying...Chaotic good characters are trying to avoid collisions... but wait...the chaotic neutrals are aswell...
I'm sorry you completely lost me with your examples.
An easier solution is just, increase the distance the higher the probability of collision, including possible re-routing.
> It was slow. It was cautious. It was safe, a more attentive driver than any human could possibly ever be.
> And later that evening, as I snaked my way back to San Francisco on the four-to-five lane wide Interstate 880, I wished it was the Google car driving me home.
...
> That's a terrible complaint, isn't it? The impeccable standards are a sign the technology is really quite magnificent.
...
> The way Google sees it is that even with those theoretical scenarios - where it's forced to choose between what it thinks is the least harmful - the Google car is still far smarter than a human, and, crucially, will save thousands of lives by existing in the first place.
> Riding in a Google car may be slow and boring, but the effect it will have on the world certainly isn't.
In fact, look at self-driving taxis or cars as your personal and private "bus" or limo, if you will.
Don't know why they didn't do mass transportation, but my guess is legislation and unions. Getting rid of personal cars and human driving taxis is private industry. Mass transportation will have a large percentage be public sector workers and union workers.
If this is the path of least resistence and is proven to work, they will most likely go after mass transportation as well.
Buses are more complicated than cars (city driving, people at stops etc), so doing cars first makes sense.
>Buses are more complicated than cars (city driving, people at stops etc), so doing cars first makes sense.
I don't think this is a valid argument. I see self driving cars operating in Mountain View (a city, albeit not a crazy driving city like SF or Boston), these cars are aware of people as stated in the article, the condition of people at stops can start with stopping at... well stops ->recognizing people that are waiting there. Maybe prepayment to figure out how many people should have boarded.
That's certainly where this is going, but the problem with mass transit is bootstrapping.
Currently, I take one of the Amtrak trains regularly, not terribly fast, but I'm not driving. However, I still need a car at either end of the train for the short hop to house/office.
If I could summon a self-driving car, I would take the train more often. In addition, if I could summon a self-driving car, I wouldn't worry about keeping a strict schedule at the office (very few places have trains/subways that run with enough frequency that you don't have to pay attention to schedule).
This might also be a good intro for fuel efficient vehicles and alternative fuels: human drivers demand acceleration and avoid slower vehicles (never mind that 80% of their time will be going 13mph in traffic) but someone who's teleworking, playing a game or watching Netflix won't care since they'll get to work at the same time either way.
Classic case is finance - people derided Canada's banking sector as being boring and not taking advantage of new fancy financial instruments, but lo and behold the financial recession hits and Canada's banks remain insulated from what happened in the US, Ireland, and other places that were exciting and flashy. [1]
This isn't to say that all of life should be boring and conservative, just that we shouldn't chase 'interesting' at the expense of 'stable and functional' in core areas of our society (or our apps).
[1] http://policyoptions.irpp.org/issues/the-fault-lines-of-fede...
What if you were being pursued by an unknown organization, could the driverless car perform evasion driving techniques and safely get it's occupant out of harms way?
There's so many edge cases.
Fixed that for you :) Although I think you're joking, coming up with scenarios like these is often exactly what people do when change is presented to them - they fill in the gaps with their imagination. Our imagination can create infinite edge cases, but its bad at predicting frequency. Just remove 'driverless car' from your statement and ask yourself the same thing. How would you get out of that?
well, my point was that stuff we consider 'hollywood' happens day to day in less wealthy parts of the world, that's what I meant, it'd be interesting to see what solutions we come up with.
And to answer somebody who said, you couldn't get out of a situation yourself, that's the whole point of trusting a driverless AI, I want it to do the impossible for me.
Imagine if you had driverless cars in the favelas, that's what I mean.
Like that simpsons episode, where the truck automatically evades danger. I wonder if it's possible to have a driverless car that is able to perform these type of 'humanly impossible' feats.
I don't see why not, a sensor monitoring the world in ultra slow motion, able to steer it's occupants to safety when it detects a threat.
You realize you're debunking one fantasy with another, right?
In fact, it's unlikely that the manufacturer wouldn't do it directly since they're going to have to survive the first wave of lawsuits any time someone gets in a crash and tries to claim it was a software bug. We've already seen this with Google's use of footage to demonstrate that crashes were caused by the other driver – I find it hard to believe they'd risk losing a lawsuit over a few dollars worth of flash memory?
Do you find it hard to believe a car manufacturer would risk losing far more than a mere lawsuit by cheating emissions tests, despite the inevitability of getting caught? Or would design their cars in such a way as to allow the brakes to be tampered with over an open internet connection? Or would willingly allow cars on the market with a fatal mechanical flaw, because the calculus tells them paying out death settlements would cost less than a recall and redesign? Auto manufacturers consider the cost of human life and breaking the law as a matter of course, and there is always an acceptable amount of death and illegality they're willing to risk.
Nothing about this scenario requires autonomous cars anyway. They could already be packed with cameras, and streaming panoramic video to the police, yet they aren't. That 'few dollars of flash memory' would have to scale up to a few million dollars or more of something far more persistent and regulated to even be relevant - otherwise the volatility of the memory itself becomes a legal liability. It would also eat bandwidth and require infrastructure and maintenance, all of which would add to the cost of the vehicle, and subtract from its profitability.
While it may be a plausible attempt at futurism, it is far from being certain.
That's an unrelated discussion. We're talking about whether to add an additional benefit to work which they already need to do and which is both obviously a simple addition and a benefit to both the manufacturer and purchaser, not to mention the their insurance company.
> Nothing about this scenario requires autonomous cars anyway. They could already be packed with cameras, and streaming panoramic video to the police, yet they aren't.
They haven't been in the past because those would be new, otherwise unnecessary equipment. You're forgetting that any self-driving car must have the cameras built in to work at all, which means that instead of adding a bunch of cameras you're only adding enough storage to retain footage for some time period.
Similarly, the cost of adding a cellular connection might deter some people if this was the only use but since that's rapidly becoming a common feature you're really only adding the cost of using some extra bandwidth when someone hits a panic button and it's unlikely that anyone is going to begrudge the $0.02 that'll cost in any scenario where this would be relevant.
It does not apply to the google car since the way I understand it their maximum emergency level is "full stop", but it's something that would probably come up at some point, and I'm not sure what the "right" answer is, if there is even one.
(In any case, the "correct" answer would be "Why the F--- should I be in such a situation, and what should I have done differently to avoid being in such a situation?")
Honestly, why are people so fascinated with these never-happens-in-a-lifetime ethical thought experiments when it comes to self-driving cars? We people aren't that good with ethical problems either, and we have no problem letting people drive.
At long last we'll have "auto"-mobiles!
Interestingly, a big influence on the development of GSV was Sebastian Thrun, who worked on it in between two self-driving car projects.
I think a truly revolutionary driverless technology would be if you could buy sensors and a DIY kit from walmart, hook it up to your car to turn it into a driverless car.
A pickup truck from 1970s could be soon be driverless. Let a human drive when he feel like it but let the computer take over in dangerous and emergency situations where the human mind would be overloaded with sensory information and slow reaction speeds and lack of rational thinking could cause danger.
It reminds me of the new recovery systems being implemented in F-16s, when the pilot is passed out and plane sees imminent terrain danger, it will perform a recovery maneuver automatically.
This is the future, where AI does every possible thing for us. Even thinking. Every whim, every desires, every material can be experienced virtually via haptic controls, androids, 3d printed food of any taste and texture. We live in truly exciting times.
Think of Netflix...