Hacker News Reader

Yeah, I'm more sceptical about fully autonomous vehicles than most people, but the arguments presented in this article are bad

Even if fully autonomous cars remain impractical for real world commercial use cases because it can't match human capability to handle edge cases the investment in a programme pay off in terms of driver aids. And the fun of hands on a wheel can always be had on a racetrack.

But the goal of autonomous driving isn't "zero deaths" it's "sell cars", and convenience gets mentioned as least as often as safety as a selling point by its promoters. Since the errors autonomous vehicles make are different from those made by humans (and there's no reason to believe they will ever be zero), there's a strong argument that accidents can be further reduced by retaining an alert, responsible human driver, even if that driver's slightly less alert and responsible than they otherwise would be. That's been the case for Waymo's programmes so far (there's certainly no safety argument for making their backup drivers remote at this stage; that's all about a public show of confidence). The safety/convenience tradeoff isn't straightforward, because perfectly convenient fully autonomous vehicles means a safety benefit from fewer miles with inadequate or even drunk drivers having access to controls, but you probably wouldn't shoot for full autonomy and no controls if safety was the real goal. Frankly, even if the responsible human driver did absolutely nothing useful they'd probably net reduce road deaths simply by ensuring self driving technology doesn't massively inflate the number of miles driven per capita

> You still have to stay on the roads and use a seatbelt and use your turn signals 100% of the time, etc.

No, you don't. If you're not on a road, there's no need for any of those. You don't even need a license. Should those usecases be taken away?

> They would be building self driving cars if they cared about safety.

Imagine, for a moment, a car with all of that data that it collects distilled down into a HUD for the driver. Can you honestly tell me that it wouldn't be better than the status quo? Should we not ask for such a thing, since it doesn't remove the human from the equation?

There will never be a 100% safe automatic pilot - putting off adding safety features to cars today until we get that perfect system is, to me, really quite silly.

> No, you don't. If you're not on a road, there's no need for any of those. You don't even need a license.

I've never heard of any of this and I think you must be oversimplifying. You certainly need a driver's license even if you're going to drive on your lawn or in a parking lot or off the road somewhere, surely? I'll go do my research but there's just no way you can drive a car around private property or something without a license.

> Imagine, for a moment, a car with all of that data that it collects distilled down into a HUD for the driver. Can you honestly tell me that it wouldn't be better than the status quo?

I don't know, it sounds worse. HUDs are distracting and I would certainly be a worse driver personally if I had that in my car. I'm not sure that solution is an automatic improvement in safety.

> There will never be a 100% safe automatic pilot - putting off adding safety features to cars today until we get that perfect system is, to me, really quite silly.

Of course we shouldn't put off features for safety! No way should we do that. But if you have $50 billion to spend on the project, then no way should you not try to build a self-driving car with it. Obviously you can get a safer car with that much R&D capability if directed towards autonomy instead of just adding HUDs and better seatbelts or something.

> but there's just no way you can drive a car around private property or something without a license.

Yup, sure can. Farmers have been doing this for decades.

> HUDs are distracting and I would certainly be a worse driver personally if I had that in my car.

It's all in the implementation; consider this one: If a car is detected that is at a radically different speed from me, a red box starts to flash around it. I would certainly end up being safer, since it would let me change lanes/slow down long before my eyes and brain could detect such a speed differential. Avoidance becomes a normal maneuver, instead of an emergency maneuver.

> but if you have $50 billion to spend on the project, then no way should you not try to build a self-driving car with it.

What if you can't build a self-driving car with $50 billion? With $1 trillion? Should you just walk away while dusting your hands and going "well, we tried"?

Because that's where we are. The current efforts are still ongoing, and have been for decades. Few improvements from them are being distilled into consumer vehicles; we really need more.

Make people safer now, instead of "5 years from now".

Most people in the self-driving cars world are not talking about "taking away" other use cases for wheeled vehicles. It's a strawman.

And by saying "most people", you make a weak man argument - since you can sideline and ignore the people who are advocating for taking away manual driving options (ironically a sibling comment).

And the sibling comment is hardly a tiny minority - you can see them popping up all over the place in the comments on such articles, not to mention the Waymos and Ubers of the world.

Then say what you mean. Instead of trying to limit my freedom of owning a self-driving car, make it clear you are arguing for the right of people to drive what they want to and we’ll be on the same page.

This is the central thesis, but the author made up the idea that we are supposed to already have these cars that are safer than humans. Nobody thinks this. Nobody.

Elon Musk does

https://www.inverse.com/article/38049-elon-musk-self-driving...

What? The headline says,

> Elon Musk Says Tesla's Self-Driving Tech Already as Good as a Human Driver

Which is the exact opposite of what I said above. Elon's quote implies he thinks his cars would still kill around 30,000 people per year. That's not what I'm talking about. I'm talking about a car that's thousands of times safer than a human. Not "as safe". "As safe" as a human is terrible performance and we should not allow those cars on streets without a driver if that is the case. No way.

Sorry, I disagree.

If you want to stop 30,000 people from dying on the roads every year, don't allow multi-ton vehicles on the road.

If everyone was using motorized bikes, you would have almost no road deaths. Furthermore, people die choking on food every year. Please don't mandate us to use self-chewing teeth.

If I choke on food, that's sort of on me. If I die from a drunk driver rear-ending me at 6pm in the evening, well, that's not quite my fault now is it?

The problem is that humans are de facto terrible at driving. Cellphones and other distractions, bad decision making, emotional clouding, etc.

I've literally seen someone have to slam on their brakes and pull in the shoulder to avoid a rear end collision because they were tailgating and speeding... and then repeat the exact same mistake in less than five minutes. People are garbage at driving and when a mistake can cost not only your own life, but the lives of others, you don't have some intrinsic right to endanger me because you can't follow a posted speed limit or maintain a safe distance.

I also notice that folks who think they're the best at driving are some of the worst. "I'm a good driver! I can talk on my cellphone and drive at the same time!"

> If you want to stop 30,000 people from dying on the roads every year, don't allow multi-ton vehicles on the road.

Okay, but the idea here is to drop the deaths without impacting the economy so much that it will tank. Yes we could remove all large vehicles, but then people would starve and die anyway because they wouldn't get food to grocery stores fast enough, etc.

Yeah, but your chewing doesn't smash into my chewing because your texting.