Bolding mine
Why would you design a system that brakes but hits the obstacles as opposed to one that stops a couple of meters in front of it?
Doesn’t make sense… I mean the sensor input is the same for both, so why not do what you do one second earlier and enjoy the rest of the day?
Because of the false alarm rate. Every time it panic stops for no reason and you get rear-ended you will be … displeased … with its performance. Plus of course the vagaries of tire condition, brake condition, road condition, etc. When to last-chance brake on dry pavement and when to do the same on ice are two different things.
There are of course accidents that are unavoidable: shit just happens faster than even instant computation and race-car capable braking & cornering can overcome. So crashing is sometime inevitable. Them advertising that fact as part and parcel of the system is wise. Unlike Elmo’s “autopilot” claims that actively encourage misunderstanding what you’re dealing with.
Driving with this system for awhile now I see that it is nervous, stupid, and nearly blind compared to a highly skilled human driver. If it was scared too, it would be unusable.
You’re (unwittingly) proposing making it a very scared system. Not a winner IMO.
The purpose of an automated emergency braking system is to remove as much energy as possible from the collision, not necessarily to avoid the collision. Avoiding one is great, and if the car stops before any damage, that is wonderful, but any energy that becomes heat in the brakes or tires is energy not transmitted to a person.
I feel like for the longest time, I only heard these called “collision avoidance systems” (and automatic braking systems), but then I heard at least one manufacturer label it a “collision mitigation system”. That seems more realistic, but also legally smarter.
Reading stuff like this makes me happier than ever that I’m driving an 18-year-old Scion xB with none of that advanced stuff.
Is it even possible to buy a new car these days that isn’t overloaded with such bells and whistles?
With the latest FSD versions, what you see on the screen is completely unrelated to the internal model. There’s a separate model just for visualizations. I actually kinda wonder if they’ll get rid of it someday since it’s arguably misleading (though hopefully people aren’t actually making driving decisions based on what’s on the screen).
For better or worse, FSD 12+ is a black box on the inside. Maybe they have some ability to probe it, but it doesn’t produce a nice annotated list of nearby objects. The output is just steering/acceleration/braking/turn signal inputs.
LoFi Robotaxi Hip Hop Radio 
Waymo Depot Shenanigans To Relax/Study To
Kinda funny. Someone set up a webcam to keep track of a Waymo parking lot (with relaxing LoFi beats to study to). Sometimes they get completely jammed up and honk at each other. But even when things are going smoothly, they… aren’t very good at parking. Just all over the place for some reason, making 3-point turns for no reason, etc.
Thanks for the edumacation.
This. My impression as a passenger (who’s never read the Tesla manual) is that I was using it to compare what it “saw” to what I saw to better anticipate situations where I might have to intervene soon because I knew something it evidently did not. Like my example bicyclists.
Now learning that it’s just fanservice, I hope they turn it off. Actual parking cameras to assist the human in seeing a more complete view around their car while parking manually are real useful. I would not want a car without them anymore. But this thing?
Remember that your trusty Toaster is surrounded by cars semi-self-driving like this.
What I have found is that my own state of mind greatly affects how I perceive the car to be driving.
If I am antsy / in a hurry / busy, the car’s combo of nervousness and timidity drives me batty. If I’m in a mellow mood and sorta want to watch the scenery and listen to the tunes, not just get there ASAP Og-dammit!, the car’s driving is much more tolerable.
Our traffic here is not too dense, but it is aggressive. You can either zone out and let the mayhem flow around you, or get in there, drive like you mean it, and slay them all. The car can do neither on its own very well.
isn’t that weird/counterproductive/stupid? … providing the pilot information on which NOT to act?
It’s not useful. It’s just there to… look cool, I guess.
Early on, the visualization would be really bad, with cars spinning around in place, flashing in and out, etc. It was obviously not useful for anything, and even then didn’t really represent what the car acted on. But it looked cool when it was working. It’s much better now, but still doesn’t represent anything.
I mostly agree that they should just rip it out, but I don’t have a good idea of what they should replace it with. The third-person car view is part of the aesthetic, and when you’re parked serves a useful purpose (to open the trunk, frunk, charge port, etc.). And the navigation visualizations are mostly legitimate, I think, though people have reported that sometimes FSD overrides the path.
The new eye tracking code doesn’t let you look at the screen for more than a moment. You’ve gotta be looking out the windshield. I guess that’s one solution to the problem.
I think you can put your hand on the wheel like before if you want to look at the screen for a few extra seconds.
Put my hand on the wheel? Like some damn primate? No thanks.
You terrify me. Or, well, not too much since I live in a suburban/rural area with moderate to light traffic, and don’t drive much beyond my safe havens. You could offer me a winning Megabucks ticket and I wouldn’t drive into Boston.
It’s somewhat related to the driving model in that they’re both based on the what the car is sensing. There should be some intuitive way to let the human driver know what the car thinks it knows.
For me, the display is useful for awareness of what cars are nearby. When changing lanes on the highway, I’ll glance at the screen first, because it’s easiest to interpret if “anything’s there”. Of course, once the signal is on, the display switches to the appropriate camera.
I don’t use the car’s self driving, but having a detailed display helps me build trust. (I got notice this morning of a new version update and an offer to try self driving again.)
This has been making the news rounds over the last day. NHTSA has opened an investigation into Tesla’s FSD. It seems that they’re looking into whether it understands when it’s rendered incapable of proceeding on its own when the camera-only systems are hampered by the environment it’s being operated in.
14 investigations? Well, I’m glad they’re counting. Somehow none of those have resulted in any findings of note. One of them resulted in a recall… which was to make the text of the on-screen icons slightly larger. Maybe they’ll have better luck this time.
Your comment about “trust” makes sense That was certainly my own naïve reaction to the display.
And that trust would be well-founded if the software drawing the screen was the same software doing the driving.
But since they are two disconnected systems sharing only the camera feeds, any “trust” the display engenders is false trust. Maybe the driving software “sees” the car the display software shows in the lane alongside and maybe it doesn’t. No way to know.
I think the basic question of whether Tesla driving software in general “sees” the surrounding situation well enough to drive is well-settled: it does. And plenty good enough.
Whether it does so in any specific situation such as right now when / where you are? The display is uninformative (or at least unreliably informative) on that question.
I’m in Phoenix, and it seems that i can take a waymo to get from my host’s home to the airport. I’m going to give it a try. All me Tuesday how it went.
(Should i warn my husband that I’m putting my life in the hands of an autonomous vehicle?)
How much life insurance do you have? If lots, don’t tell him. Let the check be a big surprise. 