A good perspective on self-driving cars

harrisonpatm

10 kW
Joined
Aug 8, 2022
Messages
825

Thought it was a good and newer point of view on how full-self driving isn't what it says to be, and, if you watch to the end, what it could look like if it's something that we actually want. Which I don't, I think getting behind the wheel of a multi-ton vehicle and assuming it'll work fine is lunacy. But I don't mind considering what could be.

Plus, skip to 14:15 for his statistical interpretation of how FSD is far below its claims of being way safer than humans.
 
Which I don't, I think getting behind the wheel of a multi-ton vehicle and assuming it'll work fine is lunacy.
Keep in mind that you do just that when you get in an airplane. Airliners are rarely hand flown any more, and Cat III certified planes can fly from pushback to arriving at the next gate without anyone touching the flight (or ground) controls. And flying is a lot more complicated than driving.
 
Keep in mind that you do just that when you get in an airplane. Airliners are rarely hand flown any more, and Cat III certified planes can fly from pushback to arriving at the next gate without anyone touching the flight (or ground) controls. And flying is a lot more complicated than driving.
Variables aside (way less traffic in the sky than on a road, exponentially fewer planes in the sky, less frequency of air travel on a person's daily use compared to car travel, just off the top of my head)...Yeah, I don't get in airplanes either. And it's pretty well documented that overreliance on fly-by-wire systems leads to lack of pilot skill when it comes to actual emergencies or technical failures that require skilled human input.

Old article: Crash: how computers are setting us up for disaster | Tim Harford

"This makes it very hard to crash an A330, and the plane had a superb safety record: there had been no crashes in commercial service in the first 15 years after it was introduced in 1994. But, paradoxically, there is a risk to building a plane that protects pilots so assiduously from even the tiniest error. It means that when something challenging does occur, the pilots will have very little experience to draw on as they try to meet that challenge."

 
That's what simulator training is supposed to help with (since many of the hardest to handle emergencies would not be safe to learn to handle "in reality", until you already do know how to handle them). (stated from my non-aviator POV, having only followed this kind of thing via air incident reports, pilot discussions online, and "disaster analyzation" videos of various types)

Of course, that's only for pilots that have access to such training, and companies that can afford to do such training for their pilots or care to do so.
 
Variables aside (way less traffic in the sky than on a road, exponentially fewer planes in the sky, less frequency of air travel on a person's daily use compared to car travel, just off the top of my head)..
And way higher speeds, an extra dimension, the inability to stop in an emergency, the much greater complexity of aircraft, the higher workload . . .

It is worthwhile to study why aviation is so much safer than driving. It's partly due to automation and training, but a large part of it is how aircraft communicate with each other. That's something that is missing in all existing self driving systems.
 
I'm a big fan of this guy's youtube channel. And some of his music.

Absolutely no way i'd trust the current state of the art for anything critical like driving.

AI's reasoning abilities are still poor. I don't think we are anywhere near making an AI driver as accurate as a human. Sure, some piece of software can react quicker than a human, but still can't think so deeply.

Remember those Cruise robotaxis in San Francisco? right after the city pulled their license, they said that human intervention was needed every 4-5 miles.

I think overall, AI world is really overselling products that almost work as if they fully worked. The only segment of AI world i find really impressive is image generation.
 
And way higher speeds, an extra dimension, the inability to stop in an emergency, the much greater complexity of aircraft, the higher workload . . .

It is worthwhile to study why aviation is so much safer than driving. It's partly due to automation and training, but a large part of it is how aircraft communicate with each other. That's something that is missing in all existing self driving systems.
Well, I gave up my wings around the time my 2nd kiddo was born, figured I needed a less stupidly expensive hobby, like Idunno, tossing diamonds into Piranha to see em dissolve....

With that said, why is air travel so much safer?

Order of magnitude my friend, look up the total number of planes that are air cert'd (something to the order of 200k) Look at how many cars are on the road in oh, say NorCal (ok, bugger, can't get that specific) California. Just north of 14m.

Average time spent training to get an air license 15 hours to first Solo, then 1500 hours before you get your first wings.
Average time for a California Drivers license...20 min.

Now notice, I skipped ground school and any training for the drivers test, just to make it about equitable and to keep my brain from melting.

So, once you look at them numbers, do ya still feel like you need to really cognate much on why cars go crashy crash so often?

If you are still concerned, keep in mind, those plane numbers.. are total for the US, and the cars.. yep, just one state. (a populous one, but just one).

Also a pilot has on average like 3 towers in radio range (been a while and that is from the hip, not looked up) full of people whose entire job it is to keep you from bumping into something else in the big blue empty. My first like 10 flights I think I only saw planes when on approach or taking off.

Also, you have to go through a detailed set of processes every time you fly, and there are always eyes on you. Despite all of this people manage to lawn dart every year.

The question is not how it is safer, The real question is how do we have so few accidents here on terra firma?
 
And ultimately, autopilot on planes isn't what the video, or my post, was about. Hobby drones nowadays come with autopilot too: return-to-home, waypoint finding, collision avoidance. But it's silly to say "drones are self flying, why can't cars be self driving?" Because they're two completely different things.

I think overall, AI world is really overselling products that almost work as if they fully worked.
The video points out how Musk has done that every year for the past 5-6 years, promising FSD just around the corner.

It is worthwhile to study why aviation is so much safer than driving. It's partly due to automation and training, but a large part of it is how aircraft communicate with each other. That's something that is missing in all existing self driving systems.
Can't argue with that. The end of the video describes how FSD could look, he compared it to a school of fish. Right now every autopilot car acts on its own and makes its own decisions. That's not even how people drive now. Come up to a 4-way stop, and you have to look at the other drivers, communicate with them via waves or headlight flashes, make decisions based on who is going and who isn't. Merge on and off a roundabout, same thing. Airplane and drone autopilots aren't expected to do that.
 
Back
Top