I’m also confused, since there seem to be a lot of smart people thinking this is just around the corner, but I didn’t think AI was nearly good enough to pass this minimal test:
Can recognize both of, tell the difference between, and react appropriately to, a police officer standing with her hand palm out, and an office standing and waving her hand in a 'OK, go" motion?
Are there cars that are close to being able to do this?
We’re not going to get full acceptance of self-driving cars from AI learning software in it’s current state. The public loves technology these days, but just in the early stages of development, with a lack of specific law to cover self-driving cars, the experiments will soon lead to tragic loss of human life and a great deal of law restricting the use of self driving vehicles.
However, I think it might only take about 10 years to get a self-driving car ready that could meet reasonable safety standards, but probably more time than that to weave through the legal and political minefield.
At present, autonomous cars like Tesla and Cadillac are pretty much at Level 3: it can drive itself under some limited circumstances, but the driver is still required to pay attention and must be ready to intervene at any time.
Yes, thank you for the link and the definitions. It appears I’m talking about a Level 4 car, but the devil is in the definition of the “Operational Design Domain.” I’m sure there are Level 4 vehicles now, such as the automated truck train, that just have a limited ODD.
I wonder whether that’s true. The Rio Tinto/Pilbara article is about a real train, as in “steel wheels traveling down a railroad,” not a concatenation of 53-foot trailers towed behind a steerable tractor piloted along a paved surface.
To my knowledge, at this time there are no Level 4 autonomous road-going vehicles available for sale to, or use by, the general public.
Self-driving cars without networking is a dead end. We’ll just end up in massive self-driving gridlock if the cars can’t communicate, and we’ll lose a major benefit of self-driving cars which is to clear up the massive traffic problems in metropolitan areas. Hacking can be prevented, but since we’ve accepted shoddy protection in consumer products so far, and don’t hold suppliers liable for allowing hacking in the first place, it’s probably not going to change when it comes to cars.
Non-self driving cars will be networked also. Cars will communicate with nearby cars to keep their distance and automatically react to upcoming traffic situations.
If all work on self-driving cars stopped today, the hacking problem would be just as bad.
Remind me never to drive a car in California again. I figured I’d be too old to care when the robots took over, but I see they won’t be facing that much resistance in the first place, they won’t be needing terminator style functionality. Now I’m surprised we survived the blender with a brain back in the 70s.
Indeed the [NHTSA](file:///Users/donseidman/Downloads/V2V_NPRM_Fact_Sheet_121316_v1.pdf) has already “issued a Notice of Proposed Rulemaking (NPRM) to mandate vehicle-tovehicle (V2V) communication technology for new light
vehicles in the United States.” Cybersecurity is a definite item of concern they have.
It is also true that the V2V input functions as one of many overlapping sources of data about what is going on around the vehicle. Input from any source that is discordant with input from the other sources is unlikely to be acted on in a drastic fashion. Car rapidly braking ahead via the V2V channel may cause some instant slowing down and only more complete braking if the information is confirmed via other channels, for example.
Certainly hacks that get past security systems and that somehow override other inputs are not impossible. OTOH deaths that occur every day from accidents that these systems will routinely prevent are a certainty.
To be sure though the full benefits will take some time to realize if for no other reason the simple fact that the cars on the road now will last a long time. I do also suspect that a requirement for cars to broadcast safety related information will precede any requirement for cars to be able to receive such information.
Devices like that and warns that merely warn a driver of hazards may be what the report meant by
Pertinent to this thread, they agree that V2V complements autonomous systems:
And yes having cars so enabled paves the way for autonomous vehicle adoption.
Yeah I’m familiar with road trains. Snowboarder’s link was to a literal train though, one that runs on rails.
The average age of cars on the road in the US is about 11.5 years. Even if every single new car on the market from tomorrow on was self driving, it’s going to take over a decade before you see 50% penetration in to the market. Given this is clearly not the case, even in the best of circumstances, it will be at least 15 years before there is anything close to 50%. 75% will be much longer.
Aye, a freight train. With no humans. That can successfully cross hundreds of kilometers. Did I mention that there are no humans involved?
You know what else is becoming largely autos (or drones)? Every other physical aspect of mining.
I was just using it to illustrate that the argument “it’ll be decades before robots can x” are far off the mark. The early days of it are here; they are happening.
Good info; thank you. Forgive me if I respectfully disagree with your estimates, however.
I want an auto as soon as I can get one. As soon as they are available and affordable for me, I’m done driving unless absolutely necessary. I still want to get places in my comfortable little bubble, I just don’t want to have to do the actual driving, not when the auto can do it safely and efficiently. I suspect that there are a lot of people like me and that the transition will go faster than you anticipate.
Still, I’d be willing to concede that the 15 year mark should start when they actually hit the road, not today.
Yeah, but, you know, a train. It would be the most simple thing a robot has to do. It has to start and stop and that is all. If anything it is surprising it hasn’t been done earlier, particularly in outback Australia where the biggest thing that can go wrong is that it hits a kangaroo.
Well people will get cars when they can afford one and they can justify the expense. I doubt the introduction of self driving cars will make many people buy cars when they otherwise wouldn’t. More likely they will consider the self driving option when it comes time to replace their car.
The way cars are marketed suggests that lots of people enjoy driving. What percentage? I don’t know. I know I enjoy driving and although I would certainly use auto-drive in some circumstances, I would not always use it. For one, if I’m going to be sitting in a car for 30 minutes, I may as well drive it, if only to give me something to do. So for me at least, I’m not going to buy an auto when I wouldn’t already be buying a car. When it’s time to buy I will look at the self-driving option and if it is in my budget I may get it, but there are other much bigger priorities such as boot space, rear seat space for child seats, turning circle, etc. The last time I bought a car I declined to get the lane assist, parking assist, and all that because it wasn’t worth the cost.
I’d be interested to know what self driving would be worth to people. To me it’s worth about $5000. Any more and I wouldn’t bother.
I suspect there are a lot of people like me and that the transition will go slower than you anticipate ;). Honestly I don’t know. I don’t know if 10% are like me or 90% or 50%.
I’m not a luddite, I would like to see it all happen, I just think it will be a slow gradual process
$5,000 sounds about right for me too, but I imagine the value of an autonomous vehicle will really depend on your situation. Someone facing an hour-long commute through stop-start traffic with a daily battle to find a park will find a lot more value in an autonomous vehicle than someone with a pleasant 10-20 minute drive with little traffic.