Interestingly it is exactly under conditions of poor visibility that airplanes rely on their autopilot systems as instruments can tell more than the naked eye.
Long term I can imagine lane paint markers that have magnetic properties detectable through snow or other road embedded markers/emitters.
Airplanes are the exact opposite of cars–they are restricted to specific routes at specific times, and required to follow precise rules. You don’t just hop in your plane for a trip to the grocery store and the kid’s soccer game.
There are lots of big players trying to make a robo-car, and when they get it perfected, they will change society. But it’s going to take a hundred years, not 10 or 20.
What will happen when a google car gets contradictory information?
Say, a construction worker (not in uniform) flags you down to tell you that your lane is closed. Your car has to cross the solid yellow line in the road (which is, of course, illegal under normal circumstances) , and then drive in the opposite lane (illegal and suicidally dangerous!), but permissible right now because oncoming traffic is temporarily stopped by another construction worker 3 miles farther down the road. The construction workers are coordinated with walkie-talkies, but your google car can’t know that.
Sure, you could pass a law requiring construction companies to file plans with the local GPS computer network notifying it of changes in the road, just like the GPS notifies me of traffic jams today. But that law will have to specify a lot of technical stuff, and nobody yet knows how to write it.
And when your Google car runs on Windows 7 , but the GPS system runs on Windows 8.0 and the technicians are in India, and I bought my used robo-car with Windows XP installed …there are going to be serious accidents. Or more likely, the car will simply stop in place, and you will be stranded, helplessly.
I’d imagine that insurance rates would go down. I’d love to be able to sit down in the car, program in the address I’m going to and then enjoy the view or take a nap until I get there.
10 is probably optimistic. Well before 20 pretty likely for autonomous under most driving conditions but still requiring a human driver to be on board. 100? Silliness.
Contradictory information? The same way we human drivers deal with contradictory information, by having algorithms that prioritize levels. Yes, a truly completely autonomous vehicle has to be able to “see” people directing traffic and put it as a relatively higher priority bit of information, just like we do. Simple enough in a world in which some reasonable percent of vehicles are expected to be autonomous to have anyone empowered to direct traffic to have a vest/gloves with standardized symbols or even an emitter that the car can read.
You other bit is really also silliness. Amazingly for certain systems we have standards for compatibility. Gas tanks all have to be able to take the same nozzles … this really is no different.
The big question in my mind is how we transition into autonomous vehicles while dealing with a simultaneous archaic legacy system for likely several decades.
The likelihood will be through autonomous features being used as both an enhancement collision avoidance systems already available, with the system taking over when high risk is detected, and able to be turned on and off in an expansion of the role cruise control is used now but with the circumstances included as appropriate expanding to the vast majority of driving circumstances. As more are on the road then allowing for ad hoc convoys to form via V2V communication, often with special lane access. Once that critical mass hits then infrastructure supports to allow for greater support begin to occur. Cities invest in lane markings that emit to vehicles such that poor visibility due to snow (etc.) is a non-issue, so on. Cheaper for a city to enable these tightly packed ad hoc car trains to function efficiently and without accidents in bad weather on the major city highway and arterials than most other solutions available.
California will likely drive the market as the first cities to give special preferences to these sorts of vehicles will be cities like LA. The importance of the California market is huge. Probably China and European markets will not be far behind the adoption curve.
The simple and safe way to handle this, the way the current car would handle it, is the car would simple come to a smooth stop if it detected no more valid roadway. A human driver would be prompted to take control.
Yes, I would buy a Google-driven car. I will be an early adopter if I become disabled from driving otherwise. I suspect this market will be driven hard by the Babyboomers and Gen-Xers who will surely grasp at any alternative to losing the freedom of automobiles as they age.
I had already realized what a boon self-driven cars would be, but I admit it wasn’t before this thread that I realized how superior the rental or subscription model is. You could call a one person car to take you to IKEA and then enter the exact dimensions of what you need to carry to order a vehicle with the right capacity to take you and all your purchases home. The end of parking seems like an even greater gift than the end of driving.
I’d bet money that in 10 years, you’ll be able to buy a brand new car that will be able to self-drive door to door in at least 50% of the 50 most populated cities in the country. At that point it might be an expensive thing with some real limitations, kind of like buying a tesla now. I’d bet with that in 10 years you’ll still have manual controls and have to be behind the wheel and be nominally “in control”/“able to take over”, probably more due to law issues than because the car really needs your help.
Would I buy one, hell yes, I hate driving, and look forward to the fact that by the time I am nursing home age, crap eyesight/poor health won’t mean I am all but a prisoner in most parts of the country.
Don’t forget the early signers will eventually get grandfathered in to the unlimited plan, but their speed will be throttled during peak hours.
It is possible I would get a Googlemobile, I would have to check out the alternatives.
I think you many here are severely underestimating the capability of computer driven cars and vastly overestimating people’s ability to adapt to changing situations and various roads.
Most are envisioning a system that falls back to human control when it trouble, I quite see the opposite, when driving in human mode the car will take control from the person when it detects a problem, sort of like collision avoidance systems do now, but with the ability to fully recover and keep on the path.
Yep, this is big.
Absent robocars, all of us, as we age, will reach a point where giving up our car keys is the only responsible choice. I expect I’ll still be able to safely drive a car at 80, but somewhere in between 80 and 85 there will come a point.
I’m 60, so 80 is just two decades away. But I’m pretty sure we’ll have self-driving cars by then, in the world at large, and not just in specially prepared or mapped environments.
ETA: As a benchmark, think of where the Internet was, 20 years ago.
But if a robocar can handle the most difficult situations better than a person can, ISTM that it should be able to handle the prosaic ones well enough to get by.
I would.
I like the fact that Google so quickly identified the weak link in driving a car-they removed the manual steering so that the passenger couldn’t interfere with the operation. Apparently all the passenger can do is push the big red stop button. I like it.
Not necessarily.
The most difficult situations are actually fairly predictable in terms of the what to do and what the likely inputs are. We humans just don’t maintain our attention as well and don’t respond as quickly as an automated system can. The outcomes of those situations are determined by those factors. An automonous system is never a fatigued or distracted or drug/acohol impaired driver. It never forgets its training. Also , it does not get emotional about the jerk who cut in front or passed on the shoulder and it is not annoyed and impatient as a result of creeping along in heavy traffic. The lowest risk and most effficent driving decision is pretty easy to make in most problematic circumstances (if not easy to make quickly enough), so long as human failings resultant of boredom, distraction, impairment, and emotions are kept out of it.
Processing for any possible prosaic input however is a bigger challenge.
Again, the likely path of these vehicles (other than for vehicles with defined paths or territories, like buses, garbage trucks, campus vehicles, taxis, delivery vehicles, etc., and maybe semis) is through semi-automous capability. The op hypothetical proposed no price premium. That is not likely realistic. So the question may be how much more are we who look forward to having one of these sorts of cars willing to pay to have one?
It would be a totally different experience. I like driving myself for the fun of it, but I’ve enjoyed relaxing, making calls, and doing work on a laptop in the back of a car-service limo too.
If it was like riding in the back of a limo, I could see it catching on.
I think you’d need real AI to handle everything humans handle without refitting roads. And getting people to pay for that stuff via taxes is hard. Plus there’s a lot of the country that doesn’t even have the normal paved roads.
Having a sort of more advanced cruise control would be a lot better. Adding in collision detection would be good. However, it will also probably come with an inability to speed at all, since you will need to be monitored. I don’t see speed limits going up while there are still normal cars on the road.
A “real” AI, by recent definitions, would be capable of learning and performing every software programming and hardware engineering task performed by the people who built the AI in the first place. It would be capable of understanding nearly all nuances of human language, it would be capable of repairing and designing new machinery, the list goes on.
An AI that drives a car needs to recognize only very specific human signage and physical objects that may impede the cars path. It would factor in known variables for the car’s motion to plan future movements. This is a subset of a subset of a subset of the capabilities that a “real” AI would need to be capable of.
Well this may change things and speed it up some …
The point had been made earlier that vehicle to vehicle (V2V) and infrastructure to vehicle (e.g. lights information) will be a key part of the advantage as some critical mass of autonomous vehicles approaches. The biggest factor in that being that an autonomous vehicle surrounded by less predictable human driven ones has a harder task than one that is informed of what other vehicles are doing the very instant it is being done, even out of visual range.
Looks like autonomous vehicles will get the information before as many cars are under autonomous control:
From the plan:
Absolutely.
For those who are worried about a software crash (leading to a real-world crash), we already have that in human drivers: we call it “falling asleep”. Or drugs. Or drinking. Or epilepsy. Or heart attacks. Or texting. Or even just being distracted at the wrong time.
We can cut down on the amount of system crashes. We can’t seem to engineer out the bugs in humans, though. I’d much rather the other cars on the road be driven by something which is always aware, something which can’t be distracted.
One article I read made a very good point: By the time we’ve got self-driving cars, nobody will notice. We’ve already got cars that can park themselves that can put on the brakes if they see something you don’t, that can pulse the brakes for you if you lose traction. Self-driving cars won’t just suddenly arrive- they’ll get here piece-by-piece, and that process has already started. By the time we get self-driving cars, we’ll be so used to having the cars help us drive that it just won’t be that big a deal.
And man am I looking forward to being able to read a book or nap on a road trip.