Back in high school, I read a book called 2081 by Gerard K. O’Neill. It described an optimistic future of wealth and space travel. There were also self-driving cars, but the system described there had one key feature that seems to be missing from the SDC systems of today: local hard-coded road information. In complex areas, there were pegs hammered into the road every few metres that when queried by the cars gave information about curves, hills, intersections, etc in the immediate area. We would now call this RFID. This would seem to fill a missing link between GPS-type large-scale systems and local autonomous decisions. The biggest obstacle might be standardizing the pegs so that all cars can use them…
I’ve heard this notion before, but I’m very skeptical of it. Even assistive or semi-autonomous systems that cover several dozen specific kinds of situations isn’t going to represent a convergence on full autonomy. There are probably more like thousands of different situations that we humans have to passably handle while driving.
While I disagree with a great deal of what the blogger Atrios has to say about self-driving cars, this is one point he repeatedly emphasizes where I fully agree with him.
Driving is complicated, and for a computer, safety is probably the easiest challenge: it can detect where other things (buildings, other vehicles, pedestrians, hydrants, signs) are and which direction they’re moving, and make sure the car doesn’t hit any of them. (IIRC, at least one car manufacturer already has a system where the car will take control from you under certain circumstances to keep you from hitting something.) Getting from point A to point B at a reasonable speed, without getting confused along the way, is the tricky part.
What’s opaque to a layman like me is whether this problem will be easily surmounted by a couple more iterations of Moore’s Law, or whether it’s a genuinely intractable problem like ‘true AI’, or something in between.
I’d really like self-driving cars to be a reality by the time I ought to give up my car keys in ~20 years, but just because I want something badly enough doesn’t mean it’s gonna happen.
Yes, this exactly. Most people think they are good drivers but some really are and have a clean record to back it up.
Also I wonder if different makes/models of cars will have better safety? Seems like they would, maybe that will be a new marketing angle, “The safest self-driving software in the industry.”
Self driving cars are crucial, people need a safe place to text from.
Nobody needs an machine to solve quadratic equations or polynomials. Mr baggage should stop wasting our time and money on these difference engines.
I’m not sure that lack of a record is the best indicator of one’s driving ability, though. Avoiding accidents is certainly a key factor, but getting nailed speeding doesn’t necessarily make you a better driver than a Chinese who’s discourteous and blocking the left lane.
At the very least, we would have avoided auto-correct. 
As for self-driving cars, they are almost certainly going to be a real, major, important thing. Eventually. Definitely not in the next 5 years. Not even in the next 10. Oh, sure, they’ll exist, in some fashion, with a certain amount of ability. They already do. But a fully autonomous, trust-worthy automobile that will take you from (more or less) arbitrary Point A to arbitrary Point B is much further away. I will probably be able to see them on the roads before I’m too old to drive (I’m currently 35), but I’m not certain of that.
I’ve already bought an old farm and a barchetta to restore.
Other people have mentioned that there is already good driver risk from being hit by bad drivers, and there is also the passenger risk. I read all the time about passengers killed by drunk or young drivers who lose control when speeding.
You are suffering from confirmation bias, since you probably see lots more bugs than most people. I did also. I also saw lots of failing chips, and people in my industry way overestimate IC failure rates because of this.
If software was as bad as you think, our whole infrastructure would collapse within weeks. You need to build cars to failover gracefully when a bug happens (high end servers do this already) and keep improving.
If bad software was going to make cars fail, they’d be failing already considering the amount of code inside engines and control systems these days.
Uber is rolling out self-driving cars in Pittsburgh.
However, the cars are not autonomous. They will have a driver and an engineer in the front seat to take over the driving if necessary.
Ideally, all the cars will be communicating with each other, so if one does have a sensor failure or whatever it could use the information from others to stop safely, or even if the technology is sufficiently good to get to it’s destination, or somewhere it can be fixed.
That is probably a long way off, but doesn’t seem impossible.
People love to drive and the freedom and control over driving. Such people will never hand it over to AI even if AI was way better.
I guess some small futuristic people will embrace this technology. But I just see AI self driving cars as fad in media and nothing more.
Look at the computers and smartphones like Siri you can talk to now like star trek computer!! How many people you see talking into their computer or smartphoen!!:eek::eek::eek::eek:
The futuristic people will embrace this technology just like self driving cars. It fad the media is running with now self driving cars, siri, thin laptop, thin smartphone, thin and self driving cars taking over.
65% of smartphone owners using voice assistances in 2015
“Much of the growth in mobile assistant usage stems from advances in that space, including improved technology, more accurate voice recognition, raised awareness of these products.”
Stranger
As Stranger’s link shows, most of them, at least as far as smartphones are concerned. Have you really not seen many people use this feature? :eek::eek::eek::eek:
Seriously, I talk to my phone and car to call people. I talk to my phone to find nearby gas stations, make schedules, dictate text messages, tell me the weather, set my thermostat, etc. And I think I’m particularly behind the curve on this compared with other people I know. Christ, my girlfriend in high school back in 1992 talked to her carphone (okay, in a very limited way, just to call people. I thought it was the coolest thing at the time.)
Some people love to drive. For lots of others it’s an annoying necessity both to own and to operate a car. Many lovers of driving may change their minds when the cost of getting a license and insurance for manual car driving becomes significantly more expensive than just catching a ride with an autonomous vehicle, not to mention kids growing up in a completely changed world.
I think the main problem with the “reaction” to automated car incidents isn’t really that people will expect them to be perfect, but that the automated cars are going to make errors that are much different in kind to humans.
It’s hard for us to accept that even if AI is better than a person at a task, when it fails, it fails in ways that humans wouldn’t (and vice versa). It seems “dumb” or “flawed” in some way when an automated system misses something “obvious” to us.
This is what drivers taking the wheel is partially needed for, of course, but even then there’s always the chance of a car failing at an interval too fast for a human to react, but were caused due to a series of barely noticeable compounding errors a human wouldn’t have made.
I still think they’ll win out eventually, but I predict most of the backlash will be about “we can’t trust these things because they miss <something obvious to humans but hard to computers>”
And the cost of getting a license will go up as manual driving stops being seen as a necessity. Right now, it’s hard to increase the hurdles to getting a license or enforce driving laws more aggressively because the average person is a driver. When/if driving becomes some quirky niche hobby, nobody is going to care if you get your license taken away because someone else’s self-driving car reported you for cutting it off, or if you have to take a $5000 continuing education course every four years.
The development is happening a lot faster than I ever imagined it would. I think Google has the right approach, cutting the human driver out entirely. Not that people can’t handle situations computers (now) can’t handle, but the concept of handing the reins to a person in the middle of an emergent situation is a ridiculous concept.
The entire purpose of self driving cars is to take the work of driving away from the human, not to require the person to do literally everything required of driving, eyes on the road - hands on the wheel- feet at the ready, while not actually controlling the car at all.
I see self driving vehicles as a watershed event in transportation, equivalent to the creation of the locomotive or cars themselves. It has the potential to change everything about the way we use transportation.
IMO you’re half-right.
Expecting people to be passengers who’re ready to instantly be emergency drivers is silly, and a design for guaranteed failure. As you say.
But there are lots of situations where a truly driverless (i.e. un-manually-drivable) car is untenable. Parking in a large grass field at an outdoor concert or fair. Dealing with driving around a rural accident where you need to pull onto the sloping drainage shoulder in the dirt to pass. Maneuvering 4 of them around a crowded repair garage. Etc.
Those are situations where a Googlemobile will slow to a stop and sit there.
A Better-mobile would slow to a stop and the driver would activate the seldom-used manual controls to deal with the situation. Once back within the car’s comfort zone (which encompasses all but 10 minutes a year of 90% of the populace’s driving), it can be switched back to automatic and the driver can go back to [del]watching porn[/del] reading Tolstoy.
The manual controls may be buttons & sliders on a touchscreen, not a traditional wheel & pedals. But IMO they need to be someplace. Otherwise you’ve built an urban-suburban commute-mobile that, like the short ranged plug-in EVs, require their owners to own a spare car for their other missions. Which is a gigantic economic obstacle to adoption.