An autonomous car in SF blocked a fire truck responding to an emergency • Hacker News Board

Re: An autonomous car in SF blocked a fire truck responding to an emergency

#111

Post by mwint » Fri, May 27, 2022, 9:30 PM UTC

SoftTalker wrote: post #103

gojomo wrote: post #6

Given that often human-driven, or human-parked, cars create similar temporary obstacles, the most important question here is: does this happen more-often, or for longer-periods, with autonomous cars? I don't see the article, or quoted sources, even trying to make that comparison - so this is really only a half-story, compared to what's relevant. Further, given the remote-guidance possibilities with autonomous cars, i…

> an automated system could broadcast the planned-routes of dispatched priority vehicles to every autonomous car in the city, allowing the autonomous cars to preemptively clear paths, before it even becomes an issue of local-reasoning about an exceptional-situation These cars can barely drive themselves with hyper-accurate maps on a sunny day without any surprises, and we want them to receive and evaluate emergency d…

That central system is already far more real than you would think. I was amazed to download the PulsePoint app and immediately see all the fire/EMS events in real-time, with data on which units are responding/there/leaving. It’s not a huge stretch.

Re: An autonomous car in SF blocked a fire truck responding to an emergency

#112

Post by robbrown451 » Fri, May 27, 2022, 9:30 PM UTC

anyfoo wrote: post #84

robbrown451 wrote: post #77

Earlier quoted context omitted.

"But it does not change that the autonomous machine does not understand the complex world it is driving in in the slightest, and that, for example, swerving into an unmanned fruit stand without being able to brake is much better than swerving into an unmanned gas pump." It seems it would have a way of prioritizing such things. That doesn't seem particularly complicated, to be honest... weighted decision making is cer…

> It seems it would have a way of prioritizing such things. You are putting the cart before the horse. The problem is not in prioritization, the problem is in having the correct ontology to even get to the "prioritization" stage. Does the car know what a fruit stand is? Does it know what a gas pump is? Does it know how the fruit stand relates to the gas pump in "expected outcome when being hit by a car"? If you say "…

I would imagine the car has ways of identifying things that are specific hazards, such as a gas pump. Fruit stand is probably categorized as "other."

"the problem is in having the correct ontology to even get to the "prioritization" stage."

That part isn't done by the program, it is done by whoever enters the prioritization numbers. That is, someone, possibly a committee, can dial up the "avoid gas pumps" weighting relative to the "avoid baby stroller" weighting if they are concerned that cars might swerve so widely to avoid coming near a stroller than they are risking hitting a different hazard. Or they can dial up the weight of grocery carts relative to dogs, since children might be in a grocery cart. Etc.

Those are humans, who can do whatever ontological analysis they need when deciding on the the settings. The car doesn't need to access any of that, it just needs a general look up table that can help make optimal decisions based on the human-entered value system.

Re: An autonomous car in SF blocked a fire truck responding to an emergency

#113

Post by 2fast4you » Fri, May 27, 2022, 9:31 PM UTC

robbrown451 wrote: post #77

anyfoo wrote: post #45

Earlier quoted context omitted.

The problem is that there are literally uncountably many situations that a human with "general intelligence" will understand and react to accordingly. Sometimes smoothly, sometimes less so. But a non-conscious automatic entity needs to have the required behavior programmed in explicitly. So yes, you might argue that for this particular situation, you "just" need to put in the proper programming and AI/ML training and…

"But it does not change that the autonomous machine does not understand the complex world it is driving in in the slightest, and that, for example, swerving into an unmanned fruit stand without being able to brake is much better than swerving into an unmanned gas pump." It seems it would have a way of prioritizing such things. That doesn't seem particularly complicated, to be honest... weighted decision making is cer…

You're asking the car to make moral decisions. Given the choice between hitting a child or and old woman, which will it choose? A bicyclist vs a pedestrian? A lemonade vendor vs a hotdog vendor?

This is all assuming it can distinguish between all of these objects, and that a real person could assign relative moral values to hitting one over the other.

Re: An autonomous car in SF blocked a fire truck responding to an emergency

#114

Post by piperswe » Fri, May 27, 2022, 9:32 PM UTC

anyfactor wrote: post #23

The trend of autonomous cars is one of the things that I believe is not practical. Even the most sophisticated machine learning algorithm requires repetition and constant input to be optimized. But there is infinitely many variables when it comes to driving on a street level. Driving is intuitive on a city street. The scope of autonomous vehicles should be limited to places where there is clarity of transit and very…

"Places where there is clarity of transit and very few variables" sounds like railroads. High-speed rail is, in my opinion, a much more equitable and efficient use of infrastructure than letting private vehicles only owned by those who can afford to buy them fill up a private lane.

Re: An autonomous car in SF blocked a fire truck responding to an emergency

#115

Post by feoren » Fri, May 27, 2022, 9:32 PM UTC

bowsamic wrote: post #88

feoren wrote: post #85

Earlier quoted context omitted.

> needs to have the required behavior programmed in explicitly This is missing the entire point of ML. ML is literally defined as not having to explicitly program responses in for every situation. Needing to pull over because a fire truck has told you it's coming your way in 2 minutes is pretty easy compared with some of those other "uncountably many" situations these cars need to deal with. > the autonomous machine…

Do driverless cars use ML in the decision making procedure? > A calculator absolutely "understands" arithmetic. That is absurd and obviously false

> That is absurd and obviously false

It is actually "obviously true" unless you believe that human brains have a special metaphysical magic that makes them "more than just a system". That's literally the only alternative: human brains are magic and only humans are ever capable of "understanding". It's a vacuous definition of the word. Systems can understand things, which is good, because the human brain is nothing more.

See: Daniel Dennett's response to Searle's Chinese Room.

Re: An autonomous car in SF blocked a fire truck responding to an emergency

#116

Post by akira2501 » Fri, May 27, 2022, 9:34 PM UTC

babl-yc wrote: post #34

There are many long-term risks with the rise of self driving cars, but this doesn't seem like one of them. It was particularly novel to the fire department because it was a self driving car, but practically no different than a double parked car that a driver isn't in (which I do see from time-to-time). Long-term this seems like a very solvable problem given the ability to remotely operate the car.

> given the ability to remotely operate the car.

I think it's going to prove a challenge for these remote operators to connect to a car, with no situational awareness, and then quickly determine the correct course of action remotely.

I wonder if these companies have tested these scenarios, and their remote drivers. Do they do any testing at all? Do they do "check rides" with these remote drivers? I'd really like to know what side of this entire operation actually looks like.

Re: An autonomous car in SF blocked a fire truck responding to an emergency

#117

Post by Imnimo » Fri, May 27, 2022, 9:35 PM UTC

I feel like part of the difficulty here is that if one Cruise car responds this way, it means that every Cruise car responds this way. So whereas you might have a few dumb human drivers who can't handle this situation and occasionally run into issues, if this particular Cruise car were mass deployed, you'd run into it frequently.

On the other hand, it's possible that while the vehicle couldn't act correctly in this particular circumstance, it would act correctly in many similar circumstances, and thus the issue would still be rare.

Re: An autonomous car in SF blocked a fire truck responding to an emergency

#118

Post by feoren » Fri, May 27, 2022, 9:35 PM UTC

defen wrote: post #91

feoren wrote: post #85

Earlier quoted context omitted.

> needs to have the required behavior programmed in explicitly This is missing the entire point of ML. ML is literally defined as not having to explicitly program responses in for every situation. Needing to pull over because a fire truck has told you it's coming your way in 2 minutes is pretty easy compared with some of those other "uncountably many" situations these cars need to deal with. > the autonomous machine…

> This is missing the entire point of ML. ML is literally defined as not having to explicitly program responses in for every situation. So how do you know how the system will respond to an arbitrary situation? You could easily argue that we don't know how an arbitrary human will respond to an arbitrary situation, but we have systems in place to deal with the consequences if they handle it badly. For example, if a dri…

> So how do you know how the system will respond to an arbitrary situation?

You put them in that situation and see how they respond. If they respond badly, you keep training them until they respond better. I'm not saying it's easy, but I am saying it's exactly what autonomous-car developers been doing all this time.

Re: An autonomous car in SF blocked a fire truck responding to an emergency

#119

Post by klyrs » Fri, May 27, 2022, 9:35 PM UTC

gojomo wrote: post #6

Given that often human-driven, or human-parked, cars create similar temporary obstacles, the most important question here is: does this happen more-often, or for longer-periods, with autonomous cars? I don't see the article, or quoted sources, even trying to make that comparison - so this is really only a half-story, compared to what's relevant. Further, given the remote-guidance possibilities with autonomous cars, i…

There was a double-parked garbage truck in the opposing lane of traffic. What makes this news is that the other party to the blockade was an autonomous car.

Re: An autonomous car in SF blocked a fire truck responding to an emergency

#120

Post by krisoft » Fri, May 27, 2022, 9:38 PM UTC

anyfactor wrote: post #23

The trend of autonomous cars is one of the things that I believe is not practical. Even the most sophisticated machine learning algorithm requires repetition and constant input to be optimized. But there is infinitely many variables when it comes to driving on a street level. Driving is intuitive on a city street. The scope of autonomous vehicles should be limited to places where there is clarity of transit and very…

> Even the most sophisticated machine learning algorithm requires repetition and constant input to be optimized.

Self driving cars are not machine learning algorithms. They have machine learning parts in their programing but nobody reputable would hook up a neural network to the pedals and the steering and just let it rip.

You simply have a bad mental model of how these machines are built and it shows.

There are three big questions every self driving car has to answer: Where I am? What is around me? What to do next?

To reliably answer the “where” question you don’t just trust one sensor. You use multiple sensors. Yes, that includes a gps, but also cameras and a lidar and a radar too. It is not machine learning. If you are interested how this is done you need to look up topics like iterative closest point matching, multi-view geometry, kalman-filtering, bag-of-visual-words representation, and bayesian reasoning.

To reliably answer what is around the car you likewise use all of your sensors. Yes, this has bits of machine learning in it, but it is not only machine learning. You can and do a lot of model-free perception. For example if your laser is bouncing back from somewhere then there is something there. There is a lot of published literature on sensor fusion and tracking and prediction. These algorithms are not magic, and they are not machine learning. They are just what you also would come up with if you would think about the problem hard for 8 hours a day for tens of years.

Then there is the planning. These are not machine learning based algorithms. They usually generate a bunch of different plans, then cull the unsafe ones (for example the ones which would collide with a tracked object) and then you rate the plans according to some heuristic and choose the best, then repeat. It is not magic. There is a serious craft and engineering to it. You can read a good introduction to different approaches in Lavalle’s Planning Algorithms book if you are so interested.