4 ms·
> Level by level or straight to level five? This strikes me as a slightly strange question. Maybe it's just an excuse to present the six level system as a fram
by dahart 10y ago
> Level by level or straight to level five?
This strikes me as a slightly strange question. Maybe it's just an excuse to present the six level system as a framework for thinking about it, but a huge chunk of commodity new cars on the market this year are level 2. A small chunk of higher end cars are level 3. Teslas are arguably level 4, and we all know there are level 4 & 5 cars on the road around the bay and few other test spots. Still, the majority of cars on the road are level 1 (cruise control?) or 0. The time to decide on the question of whether we should skip the intermediate levels has passed, that ship has sailed, vehicles of all six categories are on the road as we speak.
- Ajedi32 10y agoAFAIK there are no cars yet that are _truly_ levels 4 or 5. Both Google and Tesla's cars strive towards that goal, but SAE level 4 requires that the system perform ["all aspects of the dynamic driving task, _even if a human driver does not respond appropriately to a request to intervene_"][1]. (Emphasis mine.) So until you can fall asleep at the wheel without endangering yourself or other passengers, neither of those cars are at level 4. Level 5 has even more stringent requirements, requiring ["the full-time performance by an automated driving system of all aspects of the dynamic driving task under all roadway and environmental conditions that can be managed by a human driver"][1]. So until a car can autonomously drive in snow and other less-than-ideal conditions, it's not level 5. [1]: https://www.sae.org/misc/pdfs/automated_driving.pdf https://www.sae.org/misc/pdfs/automated_driving.pdf
- Fricken 10y agoThe NHTSA originally designated a 0-4 system for classifying Autonomous vehicles. Then the SAE (engineers love adding features) added Level 5. When the NHTSA adopted the SAE system, they reworded the description of Level 5 to what you've quoted. And seriously, it's a meaningless designation. To have a driverless car that perform as well as an expert human driver in all conditions will take decades, it may require something approximating sci-fi grade AGI to pull it off. There are situations when driving that require creativity and generalist knowledge that is completely outside the bounds of what AI can do. There's no point in discussing L5 realistically, nobody is anywhere close to pulling it off.
- ghaff 10y agoI agree and, yet, you need something at least very close to that before you can have cars without anybody in them shuttling around to pick up people or deliver a pizza. (Which is why anyone who thinks they're going to be able to hail a roboUber in Manhattan within a decade is... optimistic.) I suppose you can construct really elaborate corner cases that you might give a car a pass on. And maybe there are ways to make things like construction zones easier for cars to navigate. But, yeah, for the most part, the computer is going to have to handle any driving situation a typical human would be able to.
- Fricken 10y agoIt'll be a while before the human is completely out of the loop, but Waymo has filed a patent for 'remote assistance in low confidence situations', and many of the companies pursuing robotaxis have signalled similar initiatives.
- ghaff 10y agoThe handoff speed is an issue as is the fact that the vehicle would have to recognize a problem in order to ask for help. But, yes, some sort of remote assistance system would seem to be one solution to at least certain edge cases.
- alkonaut 10y agoAdd to that the requirement to work without up to date map data, with inputs that require reading and understanding written text on road signs etc. To drive as a human in all situations you basically have to be human. You need to be able to understand the world like a human. Example: a highway has one side closed for roadwork. It's dark and it's snowing. At the start of the construction work there is a sign that says "wait at this point for escort car". The text might be partially covered in snow. When the escort car arrives it leads the queue of waiting cars past the road construction, in the oncoming lane. A human driver understands that the right thing to do is wait at the sign. If the sign isn't readable because of snow, the driver just gets out and brushes off the snow. The human driver understands which car is the escort car because of its markings and behavior. The human driver understands the reason the escort car is there is because traffic has to take turns in the same lane and there are no traffic lights. Even when a human driver encounters this for the first time, it will realize what's going on (it confused me the first time). I'm not sure it would even be possible to teach an autonomous agent to adapt to any unknown situation where the information is lacking, incomplete or conflicting.
- wrsh07 10y agoWell, Waymo's car drove a legally blind man as the only passenger: https://www.washingtonpost.com/local/trafficandcommuting/blind-man-sets-out-alone-in-googles-driverless-car/2016/12/13/f523ef42-c13d-11e6-8422-eac61c0ef74d_story.html https://www.washingtonpost.com/local/trafficandcommuting/bli... So it seems like they're pretty close to L4.
- ghaff 10y agoPart of the problem is that I'm not sure the defined levels are necessarily a particularly good framework for thinking about the level of automation going forward. -- What's a sufficient time margin for someone to take control? -- How commonly might someone need to take control? -- Can someone take control? -- How broad is the use case? It's pretty clear that we're pretty much at the point where a car can tool down at least certain highways in good weather with a (theoretically) attentive driver ready to take over if needed. What's less clear is what the intermediate stages are between that and "don't need a steering wheel in any weather on a maintained road and can park in an unmarked spot" looks like.
- jessriedel 10y agoThe question is about the development strategy, not marketplace existence. So long as each higher level is more valuable but more difficult than the next, cars will appear on the market at each level progressively in time. But the point is that, if you want to get to level 5, do you climb the ladder or build level 5 from the ground? In particular, if there is a handful of key tech advances that are necessary for level 5, and that make levels 1-4 trivial, then there's no point in climbing the ladder.
- dahart 10y agoFair enough, though the biggest hurdles are likely to be regulatory and not technical, right? A self driving car company could conceivably kill itself by solving all the technical problems and going straight to level 5 without regard to what lawmakers allow. The question has big implications... the post mentioned the follow-up question of whether we need stoplights anymore. We don't need stoplights anymore if all existing cars disappeared all at once, and all car manufacturers go straight to level 5 at the same time. We do need stoplights for the time being if for some reason that doesn't happen.
- losteverything 10y ago<the question of whether we should skip the intermediate levels has passed, ship has sailed, Do you think It's too late to bring them "back to port?" I see one endgame where state referrendums put ballot questions to citizens that ban driverless vehicles or require drivers on non-Federally funded roadways. Yeah, sensors will be added to cars (like seat belts were) and cars will be aware of other vehicles.
- dahart 10y ago> Do you think It's too late to bring them "back to port?" Probably, yes. > Yeah, sensors will be added to cars (like seat belts were) and cars will be aware of other vehicles. High end cars have had auto steering and auto throttle for at least a couple of years. One of my friends took me in a ride two years ago in a Mercedes that keeps a safe distance between the nearest car, and steers itself on curvy roads. You can set cruise control and go from Sacramento to Reno on I-80 with little intervention. I drove most of the new compact SUVs on the market over the weekend, and bought a 2017 Honda CRV. All of them have sensors to warn about nearby traffic. Half of them have some form of automated steering, usually called "Lane departure mitigation". The CRV will try to steer itself to stay in the lane. (It's not great, I wouldn't trust it for very long, but it is there.) It has "Active Cruise Control" where it follows the car in front with a preset time gap you can set. I let it drive me through the city yesterday, it comes to stop at a light as long as there's a car in front of me. (Which, incidentally makes it seem dangerous to use this way since it doesn't read stoplights.) The car has anti-collision automation, it will slam on the brakes and possibly steer if a crash is imminent, and it has "Vehicle stability assist" mean anti-roll and anti-slide measures that will take over brakes or steering or both. The CRV is a mass market car, not something high end or experimental, so yeah I think a high and growing level of automation is on the road already and will not go away. Fully driverless, I'd agree, that seems unlikely right now, and I wouldn't be surprised either if lawmakers or the public decided to make that illegal for now.
- CPLX 10y ago> vehicles of all six categories are on the road as we speak Unless something truly crazy happens, like rain, or someone directing traffic, or a truck painted a similar color as the sky. In self-driving Calvinball those things don't count apparently.
- mlvljr 10y agoJealous of Elon, right?