Uber vs. California - Robot cars not ready for roads?


#484

They all started out as people misinterpreting facts, or being mistrustful of facts, or being afraid of facts, or having too many facts thrown at them than they can process.

The document would be too long and detailed for anyone to properly digest and understand. Plus, the software is probably being rewritten on a frequent basis. Do they have to report every commit they make to the software repository? Probably, the courts are going to find the best solution.


#485

People rely on this stuff in situations where they shouldn’t. Tesla needs to do more to ensure driver attention.


#486

They could do more by removing the feature in future cars and doing a mandatory firmware update to disable the feature in existing cars.


#487

Their official line is that Autopilot enhances safety and prevents far more accidents than it causes. If that is true, they should have the telemetry data to prove it (in terms of accidents per vehicle mile).


#488

Calling the feature “autopilot” is what’s going to get them in the most trouble. The name writes checks the system can’t cash.


#489

This autonomous cars thing was a complete pipe dream. Dead before it gets out of the water.

There is no way they let these cars on the road. Musk and Uber are throwing gobs of money at a fantasy.


#490

lol ok


#491

It certainly was a pipe dream to imagine it would happen in the 20-teens. But it will happen soon enough. There are HOV lanes and the like already, and that’s how it will start, with autonomy only allowed on highways in reserved lanes. Then on all highways, then in neighborhood side streets at low speeds, then, in general.


#492

And let’s not pretend there aren’t dozens of “distracted idiot rams car into <insert_stationary_object_here>” that happen every single day and doesn’t warrant a mention.


#493

This road is close to my house and I drive it (in my Tesla) fairly often. Using Autopilot on that road is the act of a moron. No different than using cruise control, imo. It’s a windy, very busy road with lots of blind corners. There isn’t much any car can do if its operator is just stupid.


#494

I don’t recall where I got it, but I read some analysis that Tesla is a lot more lax than some other manufacturers about when and where the self-driving can be engaged and doesn’t monitor the driver’s visual attention as well as others.


#495

No kidding. Just a few days ago, the car in front of me slowed down and stopped at a pedestrian crosswalk. I braked and stopped behind him, as did the car behind me. Not so for car #4, which ran right into the car behind me at maybe 30-40 mph. Luckily the damage stopped there but I assume a robot car would have easily stopped in that situation.


#496

Humans are generally terrible at driving cars.

Our nervous systems weren’t really designed for that kind of thing. For 99.99% of human existence, the only time you ever moved faster than 10 mph, was if you fell off a cliff and were going to die.


#497

I don’t know why you think we’re terrible. You look at the number of accidents on the road compared to the number of people driving, and it’s not even most are terrible. I mean how are you defining “terrible?” Some people hit somethings or someone some of the time… same thing can be said about AI apparently.


#498

There are around 220 million drivers, and every year there are around 6 million crashes. So you have around 3% of us crashing into things, every year.

You’re right, it’s subjective in tend of what constitutes “terrible”, but generally people are easily distracted and not really good at making quick decisions and reactions at high speed.

Robots are probably better than most humans at driving at this point.


#499

Assuming those numbers are correct, it’s a big assumption that it’s 6 million different people who caused those wrecks. I suspect there are repeat offenders in there.In addition, we know how much insurance loves to play with data so the fact that countries that have such data tend to heavily charge inexperienced drivers means that there is a learning curve with driving so that number might tilt very young and also older. Then we have the driving under influence group… so when you keep adding things up, maybe humans are not inherently bad at driving so much as decision making skills aren’t at the top of the list for everyone which is no different than most things human related.

We made a decision, a good while back, as species that the cost of driving and the risk of driving did not outweigh the benefit, and there isn’t anything out there that suggests that there is anything else that can drive better than we can… yet.


#500

Not every idiotic driver moment results in a crash. I’ve seen countless incidents where a crash almost happens due to phone usage, not paying attention in general, texting, driving stupidly, etc. I’d trust computers way more than humans, once they get it optimized a bit more.


#501

That’s the yet part, and for every idiot that doesn’t’ cause a crash I’ve seen someone, sometimes a dozen or more someone’s react to avoid it. I take issue with the idea that humans are inherently terrible drivers statement… as opposed to what, the AI that drives into concrete dividers? That’s the yet part, even or when AI does do a better drive that would be AI at the top, humans next and everything else can’t even do it. I mean we invented driving, to say humans were meant to do it implies what exactly? We’re not very good at swimming either, compared to fish. Doesn’t mean we shouldn’t do it. And why did I pick swimming, because I can easily point at something that does it better than we do… it’s a real comparison.


#502

My problem is that most people think they are above average drivers and act accordingly. Those of us that realize that we are horrible drivers and refuse to take risks are at the mercy of ‘above average drivers’ weaving in and out of traffic.

It also doesn’t help how easy it is to get a driver’s license in the USA compared to other nations. I received a few lessons in the Netherlands over a summer before moving to the US again and I had a few lessons in the US afterwards, and the rigor was night and day. In the Netherlands, I was required to learn stick, start the car on a steep incline and the instructor had berated me often for not checking every blind spot. The US was basically blinkers and breaks in an automatic. Night and day.
I was able to pass my driver’s test on the first go, but I was the worst driver ever (and still believe myself a bad driver despite the lack of accidents).


#503

And all those concrete dividers are there because humans tend to drive off the road if they aren’t there.

I’m not saying people shouldn’t drive… Though, probably we shouldn’t, if i really think about it. Regardless, i drive a car.

My point is just that if we were to apply the same scrutiny to humans, that we are applying to autonomous vehicles… Then humans wouldn’t be driving. Because we are bad at it. Our reaction time is bad. We are chronically distracted. We make bad decisions all the time. We do drugs and drink booze.

We are monkeys, who evolved to walk around at few MPH.

Again, I’m just saying that if we make the requirement that an AI can never screw up, ever, then we should apply the same requirement to humans.