this post was submitted on 30 Oct 2024
548 points (89.0% liked)
Technology
58981 readers
4207 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each another!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, to ask if your bot can be added please contact us.
- Check for duplicates before posting, duplicates may be removed
Approved Bots
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
How are these people always such pathetic suckers.
Yeah this Tesla owner is dumb. wdym "we just need to train the AI to know what deer butts look like"? Tesla had radar and sonar, it didn't need to know what a deer's butt looks like because radar would've told it something was there! But they took it away because Musk had the genius idea of only using cameras for whatever reason.
I’d go even farther and say most driving is an edge case. I used 30 day trial of full self-driving and the results were eye opening. Not how it did: it was pretty much as expected, but looking at where it went wrong.
Full self driving did very well in “normal” cases, but I never realized just how much of driving was an “edge” case. Lane markers faded? No road edge but the ditch? Construction? Pothole? Debris? Other car does something they shouldn’t have? Traffic lights not aligned in front of you so it’s not clear what lane? Intersection not aligned so you can’t just go straight across? People intruding? Contradictory signs? Signs covered by tree branches? No sight line when turning?
After that experiment, it seems like “edge” cases are more common than “normal” cases when driving. Humans just handle it without thinking about it, but the car needs more work here
I grew up in Maine. Deer in the road isn’t an edge case there. It’s more like a nightly occurrence.
Same in Kansas. Was in a car that hit one in the 80s and see them often enough that I had to avoid one that was crossing a busy interstste highway last week.
Deer are the opposite of an edge case in the majority of the US.
Putting these valid points aside we're also all just taking for granted that the software would have properly identified a human under the same circumstances..... This could very easily have been a much more chilling outcome
I'm not taking that for granted. If it can't tell a solid object os in the road, I would guess that would be true for a human that is balled up or facing away as well.
It's no different in Southern Ontario where I live. Saw a semi truck plow into one, it really wasn't pretty. Another left a huge dent on my mom's car when she hit one driving at night.
Same, hit one just south of Lyndon at night.
I drove through rural Arkansas at sundown once. I've never seen so many deer in my life.
Being a run of the mill fascist (rather than those in power) is actually an incredibly submissive position, they just want strong daddies to take care of them and make the bad people go away. It takes courage to be a "snowflake liberal" by comparison
Edge cases (NOT features) are the thing that keeps them from reaching higher levels of autonomy. These level differences are like "most circumstances", "nearly all circumstances", "really all circumstances".
Since Tesla cares so much more about features, they will remain on level 2 for another very long time.
Deer on the road is an edge case that humans cannot handle well. In general every option other than hitting the deer is overall worse - which is why most insurance companies won't increase your rates if you hit a deer and file a claim for repairs.
The only way to not hit/kill hundreds of deer (thousands? I don't know the number) every year is to reduce rural speed limits to unreasonably slow speeds. Deer jump out of dark places right in front of cars all the time - the only option to avoid it that might work is either drive in the other lanes (which sometimes means into an oncoming car), or into the ditch (you have no clue what might be there - if you are lucky the car just rolls, but there could be large rocks or strong fence posts and the car stops instantly. Note that this all happens fast, you can't think you only get to react. Drivers in rural areas are taught to hit the brakes and maintain their lane.
Which the Tesla didn't do. It plowed full speed into the deer, which arguably made the collision much much worse than it could have been. I doubt the thing was programmed to maintain speed into a deer. The more likely alternative is that the FSD couldn't tell there was a deer there in the first place.
The problem is not that the deer was hit, a human driver may have done so as well. The actual issue is that the car didn't do anything to avoid hitting it. It didn't even register that the deer was there and, what's even worse, that there was an accident. It just continued on as if nothing happened.
Yeah, the automated system should be better than a human. That is the whole point of collision detection systems!
Right. I was trying to decide whether to mention that deer can be hard to spot in time. Even in the middle of the road like this, they’re non-reflective and there may be no movement to catch the eye. It’s very possible for a human to be zoning out and not notice this deer in time
But yeah, this is where we need the car to help. This is what the car should be better than human with. This is what would make ai a good tool to improve safety. If it saw the deer
that one was just standing there, yo
If tesla also used radar or other sensing systems instead of limiting themselves to only cameras then being in the dark wouldn't be an issue.
If I'm driving at dawn or dusk, when they're moving around in low light I'm extra careful when driving. I'm scanning the treeline, the sides of the road, the median etc because I know there's a decent chance I'll see them and I can slow down in case they make a run across the road. So far I've seen several hundred deer and I haven't hit any of them.
Tesla makes absolutely no provision in this regard.
This whole FSD thing is a massive failure of oversight, no car should be doing self driving without using cameras and radar and Tesla should be forced to refund the ~~suckers~~ customers who paid for this feature.
Sure, I do that too. I also have had damage because a deer I didn't see jumped out of the trees onto the road. (Though as others pointed out this case the deer was on the road with plenty of time to stop (or at least greatly slow down), but the Tesla did nothing.
Damn right. Stomp the brakes and take it to the face.