Disney intro was pretty epic too..
Some cars with autonomous driving/auto-braking features use radar sensors that will treat any metallic chips bags blowing down the road as a brick wall, it's just like how chaff works on a fighter jet 
I don't know why they don't integrate some lidar or radar as a fail safe. Pure optical systems have always seemed like a dumb idea.
Imagine the kind of juice it takes to record and post something that could get you on the wrong side of Disney IP lawyers. If Mark Rober disappears for a while then shows back up with a limp and some unexplained scars I bet I know exactly what happened.
theruleslawyer said:I don't know why they don't integrate some lidar or radar as a fail safe. Pure optical systems have always seemed like a dumb idea.
Teslas used to have ultrasonic sensors on the bumper then got rid of them. Great plan!!
theruleslawyer said:I don't know why they don't integrate some lidar or radar as a fail safe. Pure optical systems have always seemed like a dumb idea.
Because Musk thinks he is an engineer, but has no clue what that actually entails?
MSN.com: Tesla crashes into road barrier while operating on autopilot
Pretty sure the highway worker standing in the way is a mannequin.
Edit: Also, that is some of the worst lane change signage I have ever seen. Amazing that regular human controlled cars don't also crash into it.
In reply to newold_m (Forum Supporter) :
Thank you for sharing. I love Mark. I haven't seen much of his stuff lately. Great video. I too hope he has good lawyers.
JG Pasterjak said:Imagine the kind of juice it takes to record and post something that could get you on the wrong side of Disney IP lawyers. If Mark Rober disappears for a while then shows back up with a limp and some unexplained scars I bet I know exactly what happened.
He really wasn't giving away any secrets there, lots of other videos explaining/showing how those rides are laid out and how the various effects work.
I'd be more afraid of the DOJ considering how poorly Tesla came out in this test, even just showing the poor performance of the autopilot system on normal open roads.
The fake wall was straight up trolling Tesla and Elon LOL~
I love it~
I came here for the comparison test, but that bit at Disneyland was wild.
In reply to Colin Wood :
Yeah, I would watch an entire video just on the ride mapping he did at Disneyland.
In reply to JG Pasterjak :
To be fair, Space Mountain has had plenty of times when the ride has had a stop and the lights have had to come on, and pictures of that have come out.
Did anyone notice that at the absolute last second the Tesla disabled autopilot? Some people are theorizing that if the car detects a problem it can't deal with in time it turns the system off so the company can claim "autopilot was off at the time of the accident." I have no idea if that's true, but if you watch the center stack screen when he hits the "wall," the car senses the obstacle a few frames of video prior to hitting it, and shuts off the autopilot.
Edit: here's a video of just that part:
https://imgur.com/gallery/tesla-silently-disabling-autopilot-just-before-crashing-into-wall-3gyDeAp
But I've also read that any accident within 5 seconds of autopilot engaging or disengaging is counted as an "autopilot crash," here:
https://electrek.co/2025/03/17/tesla-fans-exposes-shadiness-defend-autopilot-crash/
In reply to dculberson :
There MAY be something to that, but any lawyer worth their salt would also want to know if it was on immediately prior (often mentioned in police/media reports) and how soon before it shut off.
The more mundane answer is that every one of these automated systems shut off as soon as the inputs get outside of the range of what it *thinks* it knows what to do with. Accordingly, Elon's appears to be programmed with unwarranted overconfidence in it's own awareness and abilities.
If I had to venture a guess, I'd say it might use the 'peripheral' vision area as part of 'seeing' the speed. Immediately before impact with the wall, it stops seeing that movement suddenly comes to almost a stop because the wall is now taking up the entirety of the range of vision, which no longer matches the speed inputs of the car... And thus it realizes it has no idea what is actually going on and shuts the system off.
dculberson said:Did anyone notice that at the absolute last second the Tesla disabled autopilot? Some people are theorizing that if the car detects a problem it can't deal with in time it turns the system off so the company can claim "autopilot was off at the time of the accident." I have no idea if that's true, but if you watch the center stack screen when he hits the "wall," the car senses the obstacle a few frames of video prior to hitting it, and shuts off the autopilot.
I don't buy the conspiracy theories, the idea that sub-second shutoff allows them to avoid blame is so transparent that nobody would ever buy it.
More likely this is just the software figured out that something was very much not right, didn't know what to do about it, and shut down as a result. While that's not really desirable behaviour, it's inherent to the "driver supplement" type of automation (level 3? I forget how those definitions work).
A couple things not adequate addressed in the video, IMHO, are that the level of fog and rain are unrealistically heavy, as is the sudden transition from bright sunny day into them. If you really had fog that heavy then I expect a human driver would have failed it too (or perhaps just parked the car as being undriveable).
The other thing he doesn't mention is that the chest-mounted lidar gadget is $30K, and that's not counting the back-end processing hardware. :)
When the Tesla system gets scared and can't figure out what to do, it screams at you and makes you take over, but it does not disable itself. It might even come to a stop--but the alarm is so disconcerting that I don't want to allow it to continue and find out.
These conversations are always really funny to me because of how polarizing they are. I think I get why--the central figure in the whole thing is a polarizing figure--but the systems we're seeing developed here are being developed by thousands of engineers, across many disciplines, and across a huge number of companies. It's not that guy you feel very strongly about, one way or another, sitting in an office on the roof of Giga Factory Nevada coding this thing by himself. These are teams of dedicated people bringing about absolutely miraculous and life-saving technology. All of the car makers have these systems at some level of development. But all of these systems can clamp down on the brakes when they detect an impending accident, and that can prevent an accident, or at least, it can lower the damage caused.
Guys, go drive one of these things. Go find a friend or a colleague who has an FSD Tesla (bonus points: go find a friend with one that doesn't have one with LiDAR), and you will be shocked at how far along they are. If you haven't tried FSD, or at least haven't tried it in the past 6 months, it's really amazing. I've heard the Mercedes system is quite far along, too.
In reply to codrus (Forum Supporter) :
As many missteps taken and corners cut, I wouldn't be surprised if Musk or his legal team thought that FSD/autopilot being off for 5 seconds or whatever was an adequate protection against liability.
Unfortunately we need to wait for that particular lawsuit to actually see if it is the plan
In reply to confuZion3 :
How capable are all of the systems out there? Are they all doing the same thing and trying to run at the same times? Or is one or two trying to stretch the capability for the sake of doing that?
More a general comment- given how long this has been worked on, and how much hype there has been for a really long time, it's kind of surprising that it's still so limited. I remember back in the mid 2010's seeing cars drive around the F campus, and seeing the U of Michigan's proving grounds closer to the end of that decade. Computers have gotten better, sensors have gotten better, and yet this is as far as we have come? Something is amiss. My "conspiracy" is that the code runs too slowly due to how it's being written.
In reply to alfadriver :
Its not that the code runs too slowly, its that they have to keep programming for edge cases that pop up 1 in 1,000,000 cases of someonegetting in a car. Unfortunately thats a couple hundred cases a day.
In reply to Mr_Asa :
Yup. Edge cases. But we're seeing progress made there, too.
Here's kind of a cool anecdote. I made an observation while driving around in Frederick. I was in the city, and I was curious how the car handled that kind of driving. There were pedestrians on sidewalks doing various things. The car appeared to be attempting to gauge their intent, and it was as right as I was about every encounter. What do I mean by gauging their intent, and who cares, and how would I know? Well, it mostly just kept on moving. It went through crosswalks correctly, even while turning, but when one particular pedestrian approached a curb and looked our way, the car carefully slowed to a stop, and she crossed right in front of me. This was not in a crosswalk. She was still on the sidewalk, and she had stopped there. It just observed her doing something that looked like getting ready to cross, and it just... let her go. I would have done exactly the same thing, but I didn't have to. She would have waited.
I'm not sure how the code is being written, but I bet a lot of it is being generated by machine learning tools. If that's the case, then these machines are starting to get closer to the right lines of code to provide instructions to the car for almost any encounter. Not 100%, but even you're not 100% prepared for every edge case--sometimes, this causes humans to have accidents.
In unrelated news, stock in Wyle E. Coyote enterprises has surged beating all expectations and is on track for a record gain.
I might have missed the answer but do self driving cars allow you to speed while in self driving mode?
VolvoHeretic said:I might have missed the answer but do self driving cars allow you to speed while in self driving mode?
Hah. Allow you to? Yes. In fact, if you allow Tesla's system the freedom to drive the way it wants, it almost certainly will speed. You can set an "offset" of x mph, or y% above or below the posted limit (including 0, so it follows it exactly). But the automatic speed offset lets the car basically go with the berkeley flow of traffic. You can then use the three driving profiles, "Chill", "Standard", and "Hurry" to determine how much it wants to offset that. Chill basically parks you in the right lane (or near it if there are lots of on-ramps and it wants to leave room) and hovers closer to the speed limit. It probably won't pass unless someone goes a bit below. Standard will drive like a normal driver. Hurry is... well... if you put it in Hurry mode, it'll do what you think it will.
You can also simply adjust the actual speed setting by scrolling the thumb wheel up or down. Is 73 in a 65 too fast? Just scroll it down to 65 or whatever. Want to drive 80? Scroll it up. It'll adjust as needed for turns and traffic but otherwise follow what you tell it to.
In reply to confuZion3 :
Good to know. Just tell the cops to send the ticket to Musk, he was driving.
You'll need to log in to post.
