I realized as I was posting that that my only recent experience is with iOS, and really hoped I would turn out to be wrong. It’s just such a stunningly irresponsible oversight that I’ve long presumed is nothing more than corporate CYA.
At least in Waze you can update an incorrect or missing speed limit. I’m a bit surprised (afaik) Google Maps doesn’t utilize this. I’m sure they had to be careful not to gameify speeding.
‘Full Self Driving’ is my personal nightmare.
Unimaginable how they don’t have a reliable system to ensure that the driver’s paying attention to prevent these sorts of mistakes. Driver monitoring is absolutely critical to safety.
Also worth noting: he has additional systems to indicate what the car is seeing and thinking, so he has some clue, but the car doesn’t actually indicate its intention very well at all. It’s not enough to simply make a maneuver, you need to communicate your intentions with your partner. The action cannot be the only indication of what’s happening.
I wasn’t expecting this to be perfect. But I was surprised the car did that bad. At least half of these were serious accidents if he hadn’t intervened. On a straight road with good lighting even.
Also, what kind of selfish a-hole does this sort of testing in live traffic with real people? Is the desire for “likes” and “views” that strong?
I’m curious what hais plan was if someone did pull up behind him. Wave them around while he waits for the right gap to appear?
It’s been 5 minutes since I watched this, and I’m still mad.
From my experience with JJs, if it ain’t fast it ain’t gonna work out.
I"d imagine he would just proceed with his turn. Maybe he cut out those pieces of video? Maybe he chose that intersection because few people turn left there? A problem for future historians to be sure.
Just looking at the map of where he is shows he is turning into a mostly residential area with a cluster of churches that has only one thru street, so I guess the left-turning traffic isn’t that busy (maybe on Sunday AM would be different).
There you go. Ruining my “does it have a high score board” joke.
I predict there will be a robust business in hacking the transponders to lie. As in “This car has 18 infants, 3 transplant hearts, and a super-sensitive nuke that will esplode at the slightest collision, so pick another trolley to hit.”
Related:
I continue to be mystified by this line of thinking throughout the thread. Is there anything to suggest he wouldn’t have just made the turn and continued the experiment on the next time round?
Sometimes when I’m driving I might have to adjust my GPS. If I’m on a separated lane with no one behind me, where I can see someone coming behind me for hundreds of yards away, I might take an extra few moments. But if I saw someone coming I’d be off and worry about the GPS later.
They definitely won’t solve the autonomous vehicle problem, I agree, but they have been rolled out in limited test uses and we were slated for a much wider rollout in the US in 2024 until the FCC changed policy in November of 2019. Three major auto markets have set communications standards and allocated spectrum to the task. It’s been a dead idea for decades, but it is actually stirring to life.
Hmm, so maybe in another 50 years, we’ll actually see something implemented…
You’re even more correct than you know. People who study computer vision and do AI work (ie. me) know that we are a long way from AI being able to do this safely. Even the most basic things humans do in a car are extremely difficult to get a computer to do. Tesla is pushing way way too hard on this stuff. Everyone acts like self-driving cars are imminent. We are decades away from them (in the form people imagine, such as driving you to work while you sleep). We haven’t even gotten to the hard stuff yet. We’re just following lane markings and maintaining distance. Even that is done poorly in all except ideal conditions (consistent traffic, good lighting, good weather, etc)
And in Tesla’s case, apparently not even then.
Just as someone who, in recent years, has casually perused papers on the subject, it seems pretty clear.
Which really annoys me, as I don’t drive, don’t want to, and was really hoping eventually a machine could do it for me.
Yep. It’s counterintuitive when looking from the outside. Computers are great at a lot of things humans struggle with like crunching numbers. They’re terrible at things the youngest kids can do like pointing at a dog or separate laundry.
I gotta say, I would mind my ridiculous commute a whole lot less if I could just knit my way through it.
A-yup.
I like that last one. Reminds me of the old carnival trick of “hypnotising” a chicken by quickly drawing a chalk line across its field of vision, blowing its one-watt mind.
My mind went to “does it think it’s a cat?”