Would You Own A Self-Driving Car?

[MoonFlier]

The real reason I'm for it

The reason I take the train when I need to commute.
Side note: also the best place to concentrate on a coding issue. Nothing like having to retreat into oneself while surrounded by strangers to help with the focus. Some of the most incredible code I've written has been on that train.

 

[JamieUK92]

Only if I had written the code. I don’t trust other programmers with my life.

 

[MoonFlier]

Only if I had written the code. I don’t trust other programmers with my life.

So much packed into such a short statement. All true.

1 - Trust. The theme of this whole thread really.
2 - Understanding of the technology
3 - Human nature is to fumble. We can't have any fumbles when we're rocketing down the highway at 70mph.
4 - Hacking (implied)

 

[JamieUK92]

So much packed into such a short statement. All true.

1 - Trust. The theme of this whole thread really.
2 - Understanding of the technology
3 - Human nature is to fumble. We can't have any fumbles when we're rocketing down the highway at 70mph.
4 - Hacking (implied)

Humans are stupid.

Computers do exactly what humans tell them to do.

Humans are stupid.

I’d want to write my own and be responsible for my own catastrophic death and I’d make sure only 1 model was produced lol

 

[MoonFlier]

Humans are stupid.

Computers do exactly what you tell them to do.

Humans are stupid

I’d want to write my own and be responsible for my own catastrophic death and I’d make sure only 1 model was produced lol

We are stupid. It's quite amazing at how well we've built up our knowledge base on the work of others. Most kids in school these days don't even understand how trig works or see the patterns. They just punch the numbers into a calculator.

With self driving cars will come the kit version of the same. Let the companies figure out the detection and relay devices (they're still kind of rough) and then we will be able to work the system with those inputs. This means more trouble on the road though, and it will be interesting to see how the government regulates this, I think we still have a decade to go before we see this.

Re: computers doing exactly what you tell them to do... this is why we need just a touch of AI involved. The cars need to learn and know that not everything is on or off for decisions. And an indecisive car on the road can be worse than one willing to run off the road.

 

[JamieUK92]

We are stupid. It's quite amazing at how well we've built up our knowledge base on the work of others. Most kids in school these days don't even understand how trig works or see the patterns. They just punch the numbers into a calculator.

With self driving cars will come the kit version of the same. Let the companies figure out the detection and relay devices (they're still kind of rough) and then we will be able to work the system with those inputs. This means more trouble on the road though, and it will be interesting to see how the government regulates this, I think we still have a decade to go before we see this.

Re: computers doing exactly what you tell them to do... this is why we need just a touch of AI involved. The cars need to learn and know that not everything is on or off for decisions. And an indecisive car on the road can be worse than one willing to run off the road.

It would have to have some degree of AI for sure. I don’t see how it would work without it. Even though everything with a semi complicated decision algorithm based off of priori knowledge is considered AI these days .

I’d personally only call something AI if it senses it’s own environment. In this case some form of computer vision and image processing would be the sensing of the road.

Anyway I don’t mean the human user of the system, I mean the programmer. AI is still a set of instructions that a human tells the computer to perform based off of input. Whether that input is from sensory data like a stream of images or whether it’s from human input is irrelevant.

 

[sprinkles]

I wonder how they handle animals darting into the road.

 

[JamieUK92]

I wonder how they handle animals darting into the road.

Well current computer vision and object recognition would be able to detect it but the problem is humans are much better at it. Saying that though, if it utilised some kind of camera technology that allowed it a full 360 degree view, then it could actually be more reactive than humans. We are probably quite a way off with this.

Another thing that may actually make it more safe if the tech was advanced enough. Is that human reactions vary and we are very easily distracted. A computers processing of the event and then reaction time would always be consistent. Computers don’t get distracted.

Still maybe in 50+ years I might trust the technology.

 

[sprinkles]

Well current computer vision and object recognition would be able to detect it but the problem is humans are much better at it. Saying that though, if it utilised some kind of camera technology that allowed it a full 360 degree view, then it could actually be more reactive than humans. We are probably quite a way off with this.

Another thing that may actually make it more safe if the tech was advanced enough. Is that human reactions vary and we are very easily distracted. A computers processing of the event and then reaction time would always be consistent. Computers don’t get distracted.

Still maybe in 50+ years I might trust the technology.

I'm more interested in cases where it's safer to hit the animal than avoid it. Do we teach an AI to make choices like that?

 

[JamieUK92]

I'm more interested in cases where it's safer to hit the animal than avoid it. Do we teach an AI to make choices like that?

Oh that’s a very interesting point. It’s certainly a tough problem. I guess it would simply have to be able to make these decisions, just like humans do. However, i think it ultimately boils down to how advanced the computer vision is.

Humans process imagery in so much detail and develop an almost instant mental model of what is happening. Unfortunately, our conscious mind can sometimes be elsewhere and not taking notice to the models produced by the sub conscious. Leading to poor reactions or no reaction at all.

Current computer vision algorithms are obviously no match for human vision (yet), in terms of how detailed the model of the current situation is. So, coming to a decision of what is “safer” would probably be less accurate, given the limited amount of data contained in the model. However, with computers the constant monitoring of this model is gaurenteed and so a reaction in a reasonable time is also gaurenteed. The same cannot be said of humans.

 

[sprinkles]

Oh that’s a very interesting point. It’s certainly a tough problem. I guess it would simply have to be able to make these decisions, just like humans do. However, i think it ultimately boils down to how advanced the computer vision is.

Humans process imagery in so much detail and develop an almost instant mental model of what is happening. Unfortunately, our conscious mind can sometimes be elsewhere and not taking notice to the models produced by the sub conscious. Leading to poor reactions or no reaction at all.

Current computer vision algorithms are obviously no match for human vision (yet), in terms of how detailed the model of the current situation is. So, coming to a decision of what is “safer” would probably be less accurate, given the limited amount of data contained in the model. However, with computers the constant monitoring of this model is gaurenteed and so a reaction in a reasonable time is also gaurenteed. The same cannot be said of humans.

Yeah, I remember there being a slight controversy in teaching driving AI to deal with ethical problems. For example there was a question of what to do if the AI has a choice between killing one person or killing a crowd of people. It's unlikely to ever happen because there are plenty of ways to avoid ever getting into that situation, but it's an interesting what-if.

Edit:
Also maybe an AI can act in ways that humans don't. Most fatalities involving accidents with deer are because of trying to miss the deer instead of hitting it. Also if you slam on the brakes while you hit the deer, it causes the front of the vehicle to dip which increases the chance of the deer being thrown up into the windshield, but if you release the brakes before impact the front will come up and absorb the impact more.

 

[acd]

I've actually thought about this and I'm not comfortable with automated driving because I don't like the idea of not being in control.

There are so many split second decisions to make when driving as well as ethical decisions as @sprinkles pointed out that I worry that the self driving vehicle might be limited in options. But maybe that's bs and it can be programmed with zillions of possible responses and process those faster than a human mind could.

I am sure things like this are taken into consideration by the engineers and programmers but ultimately I just don't want to relinquish my responsibility and autonomy to a machine.

I also think that ironically such advanced tech will people dumber and less self-reliant. We will devolve further into a culture that can't think for itself and relies too heavily on machines programmed by a certain section of the population. Maybe that would make individuals more vulnerable to exploitation. Look at how bots on social media have been able to influence public opinion in recent years.

 

[G4RiiM]

o hell no

 

[dragulagu]

Only if I had written the code. I don’t trust other programmers with my life.

 

[dragulagu]

In all seriousness, self-driving cars will be there sooner than later, it's going to be a long transitional period at first and starting particularly within cities.

By the way, another interesting case for self-driving vehicles is when you get into a situation where there are emergency vehicles passing by, traffic changes because you
let them pass through, but I'm not sure how the AI of a tesla handles it now.

Spoiler: Seems it's going to be there in a future update

https://interestingengineering.com/tesla-cars-set-to-detect-emergency-service-vehicles

 

[Will]

In all seriousness, self-driving cars will be there sooner than later, it's going to be a long transitional period at first and starting particularly within cities.

By the way, another interesting case for self-driving vehicles is when you get into a situation where there are emergency vehicles passing by, traffic changes because you
let them pass through, but I'm not sure how the AI of a tesla handles it now.

Spoiler: Seems it's going to be there in a future update

https://interestingengineering.com/tesla-cars-set-to-detect-emergency-service-vehicles

I can see that. Apparently, they are in the process of upgrading their computer chips and sensors to detect emergency vehicles. Though I do believe with that integration maybe during the first couple years of having that upgrade, then there will be some difficulties on identifying emergency response vehicles.

In my theory is that emergency vehicles themselves would probably be equipped with sensors or some kind of notification device letting other vehicles that it is en route. Thus they will move through traffic efficiently. In addition, they might even build an advanced monitoring system in conjunction with what's on the roads today. For example, traffic cameras might also have a detection feature to notify response teams and such with current (real time) traffic situations.

This of course can very well be controlled either by a real person or possibly in the future by A.I.

 

[sassafras]

I enjoy driving, so even when I do have the option, I'll likely be opting to drive myself.

Admittedly, I will be a little nervous on the road when these cars first enter the mainstream market.

 

[Will]

I enjoy driving, so even when I do have the option, I'll likely be opting to drive myself.

Admittedly, I will be a little nervous on the road when these cars first enter the mainstream market.

It's one of our freedom.

 

[Krypton]

I support self-driving cars being a thing, not sure I want one. Driving is fun and I want to be able to do it.

Ideally you could have a car with a self-driving mode and a breathalyzer or something. Or that hijacks your driving when you do something stupid. Second one is probably less realistic though.

 

[dragulagu]

In my theory is that emergency vehicles themselves would probably be equipped with sensors or some kind of notification device letting other vehicles that it is en route. Thus they will move through traffic efficiently. In addition, they might even build an advanced monitoring system in conjunction with what's on the roads today. For example, traffic cameras might also have a detection feature to notify response teams and such with current (real time) traffic situations.

Good theory, would actually be the most valid solution as these vehicles should always have a priority (taking into account that no accidents happen and that adding the monitoring system is economically feasible, at the least forcing a signal to the driver). Nice one dude.