Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Waymo’s Backseat Drivers: Confidential Data Reveals Self-Driving Taxi Hurdles

theinformation.com

Post reply on HN
Print view View original

71–80 of 196 posts

Re: Waymo’s Backseat Drivers: Confidential Data Reveals Self-Driving Taxi Hurdles

#71

Post by nradov »

PhasmaFelis wrote: post #68
wpietri wrote: post #62

Earlier quoted context omitted.

Yeah, I think this is going to be a long one. Rodney Brooks, AI researcher and Roomba cofounder, says he doesn't expect a real robot taxi service before 2035: https://rodneybrooks.com/my-dated-predictions/ > When Waymo can get rid of the safety drivers [...] they're getting close to something useable. I suspect that they can cheat this one a little. As long as they're in an area with good connectivity and the cars ar…

I dunno, it's hard to imagine any safe way to drive a car in traffic via cell. I don't see how you could get all-round vision with acceptable resolution, frame rate, and latency. VR might solve the viewing angle, at least, but would greatly magnify the effect of frame rate/latency drops. You don't want your driver getting VR-sick five minutes in. And that's leaving aside reliability. A 99.9% solid connection is not n…

Right. What happens when a construction crew accidentally cuts the backhaul fiber for the local cellular tower? Or if the autonomous vehicle failure occurs in a tunnel with no cell service?

Re: Waymo’s Backseat Drivers: Confidential Data Reveals Self-Driving Taxi Hurdles

#72

Post by zaroth »

zdragnar wrote: post #65
Animats wrote: post #54

That's useful info. I wonder where they had trouble with erratic steering. One Cruise video shows that where driving is alongside parked cars on a narrow street and the parking is irregular. The self driving vehicle is trying to stay in its half of the street unless absolutely necessary. Waymo still has to use "safety drivers". So far, nobody seems to have full self driving without "safety drivers". Which makes it us…

> So far, nobody seems to have full self driving without "safety drivers". A common saying around here is that we have two seasons: winter and (road) construction. Construction zones have pretty much every obstacle to automated driving you can think of: * painted lanes that don't correlate to the temporary lanes marked by cones * lanes that don't correspond to pre-programmed maps / gps * irregular and unpredictable v…

If it’s possible for remote drivers to assist a vehicle with no in-car steering apparatus through areas where the algorithm can’t go, then I think it can be done.

Getting to absolute 100% will require either AGI or an incredible infrastructure investment. Now personally I think FSD is worth on the order of $1 trillion per year to the economy, so it’s the next great Moon Shot, and totally worth every bit of infrastructure investment we can throw at it.

But it makes sense to see how much further we can get with in-car algorithmic driving before the infra investments start coming in earnest to fill in the gaps.

Another possibility is there could be ways for a passenger to assist the algorithm without actually using a steering wheel and pedals as input.

I believe the level before truly perfect FSD allows the car to get stuck as long as it does so safely. Approaching and stopping at a single lane construction zone, for instance.

The current Tesla AP does remarkably well on highways with missing lane markings. A stretch I drive every day is ground down in prep for new pavement and just has the occasional white square marking, but it’s enough for AP to lock in on. It also seems to do fine with cones.

It’s worth noting that construction zones aren’t even particularly safe for human drivers (accident rate skyrockets). So technology to make construction zones more passable overall is important, even if it just enables self driving as a side effect.

Re: Waymo’s Backseat Drivers: Confidential Data Reveals Self-Driving Taxi Hurdles

#73

Post by zaroth »

PhasmaFelis wrote: post #68
wpietri wrote: post #62

Earlier quoted context omitted.

Yeah, I think this is going to be a long one. Rodney Brooks, AI researcher and Roomba cofounder, says he doesn't expect a real robot taxi service before 2035: https://rodneybrooks.com/my-dated-predictions/ > When Waymo can get rid of the safety drivers [...] they're getting close to something useable. I suspect that they can cheat this one a little. As long as they're in an area with good connectivity and the cars ar…

I dunno, it's hard to imagine any safe way to drive a car in traffic via cell. I don't see how you could get all-round vision with acceptable resolution, frame rate, and latency. VR might solve the viewing angle, at least, but would greatly magnify the effect of frame rate/latency drops. You don't want your driver getting VR-sick five minutes in. And that's leaving aside reliability. A 99.9% solid connection is not n…

The remote assist isn’t to directly drive the vehicle like some sort of video game.

It would be to annotate something the algorithm flagged as impassable in some way such that the car can continue driving itself.

If the car entirely fails to identify a driveable lane, I don’t think you can remote in and actually active steer a human-occupied 2 ton vehicle over 4G.

Re: Waymo’s Backseat Drivers: Confidential Data Reveals Self-Driving Taxi Hurdles

#74

Post by nishantvyas »

falcor84 wrote: post #69
zdragnar wrote: post #65

Earlier quoted context omitted.

> So far, nobody seems to have full self driving without "safety drivers". A common saying around here is that we have two seasons: winter and (road) construction. Construction zones have pretty much every obstacle to automated driving you can think of: * painted lanes that don't correlate to the temporary lanes marked by cones * lanes that don't correspond to pre-programmed maps / gps * irregular and unpredictable v…

>Oh, and said protocol has to be hack-proof so trouble-makers can't start convincing cars that they're in the middle of a construction zone and force them out of their lanes on normal roads. Why would it be easier for trouble-makers to fool autonomous cars? As a human driver, I'd be fooled by pretty much any road marking or guy in an orange vest.

Correct.

It’s amazing, how much forgiving we can be for human errors ( accidents every year) but absolutely not for machine/autonomous vehicles, even when, statistically speaking, machines may make better decision much faster(or at-least no worse than human judgement)... I guess feeling/perception of being in control is more important to us...

Other interesting observation I find in every autonomous vehicle discussion is, how we only focus on edge cases... when in reality every tool that we use (including the car we been driving) today are built for general use case and operate under mostly a control environment.

Rather if we think autonomous car as additional pair of eyes and hands when we need it most might serve us well in short run before the technology gets mature over next decade or two.

I’ll be really happy and relaxed if my car can mostly (70-80%) drive it self to daily commute or next trip to LA; expecting it to be my chauffeur is bit too much, personally.

Re: Waymo’s Backseat Drivers: Confidential Data Reveals Self-Driving Taxi Hurdles

#75

Post by p1necone »

falcor84 wrote: post #69
zdragnar wrote: post #65

Earlier quoted context omitted.

> So far, nobody seems to have full self driving without "safety drivers". A common saying around here is that we have two seasons: winter and (road) construction. Construction zones have pretty much every obstacle to automated driving you can think of: * painted lanes that don't correlate to the temporary lanes marked by cones * lanes that don't correspond to pre-programmed maps / gps * irregular and unpredictable v…

>Oh, and said protocol has to be hack-proof so trouble-makers can't start convincing cars that they're in the middle of a construction zone and force them out of their lanes on normal roads. Why would it be easier for trouble-makers to fool autonomous cars? As a human driver, I'd be fooled by pretty much any road marking or guy in an orange vest.

You might be fooled by a man in an orange vest, but I doubt you'll listen to him if he's telling you to drive into oncoming traffic.

Re: Waymo’s Backseat Drivers: Confidential Data Reveals Self-Driving Taxi Hurdles

#76

Post by vidarh »

PhasmaFelis wrote: post #68
wpietri wrote: post #62

Earlier quoted context omitted.

Yeah, I think this is going to be a long one. Rodney Brooks, AI researcher and Roomba cofounder, says he doesn't expect a real robot taxi service before 2035: https://rodneybrooks.com/my-dated-predictions/ > When Waymo can get rid of the safety drivers [...] they're getting close to something useable. I suspect that they can cheat this one a little. As long as they're in an area with good connectivity and the cars ar…

I dunno, it's hard to imagine any safe way to drive a car in traffic via cell. I don't see how you could get all-round vision with acceptable resolution, frame rate, and latency. VR might solve the viewing angle, at least, but would greatly magnify the effect of frame rate/latency drops. You don't want your driver getting VR-sick five minutes in. And that's leaving aside reliability. A 99.9% solid connection is not n…

I think you misunderstand. I agree that a safety driver being expected to be able to pay attention and being ready to take over at any moment isn't viable.

But consider a point where the cars are good enough that the remaining risks are ones where they can show that they can tune the hazard detection so that it doesn't miss any potential hazard situations, but may occasionally be overly cautious, and that it can reliably and safely slow down or stop short of potential hazards it doesn't know how to handle.

In that case you might get to a point where it's ok (safety wise, if not in terms of customer satisfaction) if the car stops for 30 seconds until a human safety driver reviews the data and confirms that what the car "sees" is not a dangerous situation.

In that case you might have e.g. 10 cars per safety driver, or more, and most of the time the car might not even stop - if a driver is available to respond immediately it may be sufficient for it to slow down until it gets a response. And you can simply slowly reduce the number of safety drivers as the cars get better. For a fleet service you might well never stop having some people monitoring to respond to unexpected conditions.

Of course, for this to be viable, the car needs to be possible to be made safe without human intervention, but that safety may be achieved by opting to stop or slow down the car in situations where continuing might be perfectly safe (and with the caveat that this may e.g. restrict where it may be possible to let it drive etc.), but where the car can't yet tell by itself.

This of course presupposes specific types of failure scenarios where the car can safely find a way to come to a stop but can't safely determine if it can continue forwards. It's not a given that's achievable with low enough effort (relative to solving the issues that might cause it to fail to spot a hazard) to be worth it.

Re: Waymo’s Backseat Drivers: Confidential Data Reveals Self-Driving Taxi Hurdles

#77

Post by semi-extrinsic »

Judgmentality wrote: post #23
dumbfoundded wrote: post #18

Earlier quoted context omitted.

Level 5 driving doesn't require AGI at all. It's an extremely narrow set of problems that have to be solved incredibly well. It mostly just comes down to creating an accurate 3d representation of the world from a bunch of sensors. You also have to correctly segment and label each object in that 3d representation. If you did those two things extremely well, the actual driving logic can be hardcoded. The problem is tha…

This is incorrect. The hardest part of developing a self-driving car is predicting the world around you in the immediate future. Knowing whether or not that object is a person is a lot easier than guessing whether or not that person is going to jump out in front of the car 1 second into the future. You have to know who is going to run stop signs, when cyclists are about to cut you off, when someone is about to back u…

For me, the definite proof that we won't have self-driving anytime soon is the massive fail that was encountered in the recent chatbot fad.

One key part of driving is communicating - with pedestrians, cyclists, other drivers. This happens through body language and other fairly subtle cues.

When you can't make AI work for responding to questions given in text form on an extremely limited problem domain, how on earth would it work for something that's orders of magnitude less well defined and more broad?

Re: Waymo’s Backseat Drivers: Confidential Data Reveals Self-Driving Taxi Hurdles

#78

Post by ocdtrekkie »

DannyBee wrote: post #64
ocdtrekkie wrote: post #59

Earlier quoted context omitted.

My peeve is not the slow progress: My peeve with self-driving companies, Waymo and Tesla alike, is their constant misrepresentation of their capabilities and their timelines for the benefit of stock value and public opinion. The technology doesn't really work, and Google's marketing for the past five years has claimed regulations were the only thing holding them back.

citation? This feels like you swapped google for tesla (Elon literally said that, i can't find anything that says Google has said it)

I mean, off-hand, this paywalled article from 2016: https://www.barrons.com/articles/googles-self-driving-cars-f... The search result for it had the text "Google is so confident about its technology that the Internet search giant has already agreed to accept liability if its self-driving cars cause an accident."

Here's someone following right along with the suggestion that regulations, not lack of technology is to blame: https://www.wired.com/story/outdated-auto-safety-regulations... The author, part of the Competitive Enterprise Institute, works for Google: https://services.google.com/fh/files/misc/trade_association_...

That was page one of my search results, but suffice to say Google's been insinuating this for a while, both from the Chris Urmson era and the John Krafcik one.

I specifically referred to Tesla in my post as well. I saw the suggestion that Elon's claims about release dates for Autopilot features were effectively timed to manipulate the market. I'd give credit to that theory, or that Elon just has no clue how far he actually is from success. One of the two.

Both companies horribly misrepresent the fact that self-driving isn't around the corner.

Re: Waymo’s Backseat Drivers: Confidential Data Reveals Self-Driving Taxi Hurdles

#79

Post by beefield »

nishantvyas wrote: post #74
falcor84 wrote: post #69

Earlier quoted context omitted.

>Oh, and said protocol has to be hack-proof so trouble-makers can't start convincing cars that they're in the middle of a construction zone and force them out of their lanes on normal roads. Why would it be easier for trouble-makers to fool autonomous cars? As a human driver, I'd be fooled by pretty much any road marking or guy in an orange vest.

Correct. It’s amazing, how much forgiving we can be for human errors ( accidents every year) but absolutely not for machine/autonomous vehicles, even when, statistically speaking, machines may make better decision much faster(or at-least no worse than human judgement)... I guess feeling/perception of being in control is more important to us... Other interesting observation I find in every autonomous vehicle discussio…

This is why I wonder why platooning technology is so much less hyped. Give me a platooning hardware kit to my current car, enough users after which I can join on my longer trips and literally more than 95% of my self driving car needs are covered. I really do not care if I need to drive a 10 minute stint on the city every now and then. And If I do, I can take a taxi. But getting my hands off the wheel and eyes off the road on the highway is what would have real utility to me.

Re: Waymo’s Backseat Drivers: Confidential Data Reveals Self-Driving Taxi Hurdles

#80

Post by mdorazio »

mactrey wrote: post #33
mdorazio wrote: post #2

Interesting analysis using comments/rider scores. 10% improvement over 6 months is actually pretty good, though. Even if that rate slows as trickier edge cases get handled, It still implies early 2020s for true driverless readiness in select locations. That's pretty much on track with overall predictions for the autonomous market.

10 percentage points, 25% relative improvement (40% non-5 stars to 30%), seems quite good to me. Although I wonder if seasonality is important here - obviously Phoenix isn't Minnesota but could driving conditions have been worse in Q1 compared to this summer, from the perspective of a self-driving car?

Definitely could be a factor. Other drivers might also be more used to seeing them on the road, which could have a positive impact in some way. I think the overall trend is probably pretty strong at that level of improvement, though.
Post reply on HN