No matter what car you drive, it's still not smart enough to get you from point A to point B all on its own. Arguably, Tesla is closer to letting you safely nap behind the wheel than any other automaker, but now its vehicles are going to be less smart in the short term. Tesla has announced that it's dropping ultrasonic sensors from the Model 3 and Model Y, and its more expensive cars will lose them starting next year. And that means Tesla has to temporarily disable some Autopilot features.
[...] New Model 3 and Model Y vehicles rolling off production lines in North America, Europe, the Middle East, and Taiwan will now lack ultrasonic sensors. Enhanced camera-based Tesla Vision will take over the same tasks, but it won't be ready immediately. That means Tesla is going to have to disable several features, including Park Assist, Autopark, Summon, and Smart Summon. Tesla didn't say how much it's saving by not using the ultrasonic sensors, but it must be significant if it's willing to stop supporting some of its trademark features while it works on new computer vision tech.
Those who already have their Model 3 and Y won't have to worry about losing features, but new vehicles will be limited. Tesla says that will change in the "near future" as it works on those Tesla Vision enhancements. The features will come back in a series of over-the-air updates. Presumably, this work will be completed by the time its more expensive vehicles lose the ultrasonic sensors next year.
(Score: 4, Interesting) by helel on Tuesday October 11 2022, @01:20AM (13 children)
It feels to me like Tesla's in a death spiral. No matter how much they're saving per car it cannot possibly be worth the loss of one of their most marketable features. Autopilot already struggles to live up to their marketing hype and there's no way that taking away physical sensors is going to help it perform better. Seems like a "we lose money on every care" -> "cut corners" -> "nobody want to buy our cars" type of situation to me.
Republican Patriotism [youtube.com]
(Score: 3, Interesting) by JoeMerchant on Tuesday October 11 2022, @01:37AM (11 children)
>the loss of one of their most marketable features
But, is it a differential marketable feature? Meaning: would a Tesla customer choose to buy, or not buy a Tesla based on the presence / absence of the ultrasonic features? When Musky has promised that the affected features will be replaced with a future software update (that you may have to pay a subscription fee for)? I would guess: Tesla buyers don't really care, they're buying a Tesla either way.
We had $100 down on the Cybertruck from announcement day until a month or two ago... just so much less excited about having one at this point for some reason, and if we do eventually end up buying one I'd much rather not be first in line for the earliest of the "production" models.
Україна досі не є частиною Росії Слава Україні🌻 https://news.stanford.edu/2023/02/17/will-russia-ukraine-war-end
(Score: 2) by ElizabethGreene on Tuesday October 11 2022, @02:17AM (7 children)
I have to quietly confess I'm only holding my cybertruck preorder to see if I can flip it to someone else now. I still want an electric truck, but I used the money to pay off my house instead.
(Score: 0) by Anonymous Coward on Tuesday October 11 2022, @10:15AM (6 children)
What do you think of the Rivian truck?
(Score: 2) by FatPhil on Tuesday October 11 2022, @04:22PM
The market doesn't seem to have much faith in them: https://www.tradingview.com/symbols/NASDAQ-RIVN/
Then again, they aren't so hot on Tesla either presently: https://www.tradingview.com/symbols/NASDAQ-TSLA/
Great minds discuss ideas; average minds discuss events; small minds discuss people; the smallest discuss themselves
(Score: 3, Interesting) by ElizabethGreene on Tuesday October 11 2022, @04:24PM (4 children)
They are pretty, but I'm cheap. The only reason I was willing to spend that much on a Cybertruck was Autopilot. I'm not a great driver and understanding the limitations of driver assist systems, I thought that between the two of us we might make one decent driver.
Now, with inflation running up, the market running down, and the estimated ship dates continuing to slip I just don't see myself spending that much.
(Score: 2) by JoeMerchant on Tuesday October 11 2022, @05:39PM (3 children)
We have a 1999 pickup truck, and a 2002 sedan, both in the mid 100,000s total mileage, both kinda being patched together as we can in the absence of factory parts now.; The Cybertruck was supposed to be here by now and replace one, or both of them when they finally cost too much to repair. With the delay in launch, then adding a year or two for new launch wrinkles to smooth out, I don't think both of our 20+ year old vehicles are going to make it that long. And, $40K won't be $40K, which is a big turn-off for me. It's not that we _can't_ afford $50K for a new vehicle, it's that I just don't want to - there are other more satisfying things that the extra $20-30K over and above the alternatives could do for us.
If you're patient with your rental car office, you could rent something with driver assist for your next road trip and see if it actually lives up to your expectations. I found it... mostly annoying. Also, potentially terrifying if I ever started to rely on it. The most fun was when we were on cruise control, doing 40mph toward a small downtown area on highway 55 where the highway number was on a fresh new sign, and the speed limit 25 sign was old and dingy, and the speed-assist read the 55 sign instead of the 25 sign and started accelerating us...
Україна досі не є частиною Росії Слава Україні🌻 https://news.stanford.edu/2023/02/17/will-russia-ukraine-war-end
(Score: 2) by ElizabethGreene on Tuesday October 11 2022, @09:14PM (2 children)
I've driven a Rav4 and a Volvo with lane following assistance and both of them were well below what I've seen autopilot do. It's like comparing Tee-Ball to minor league baseball. With the pandemic that's been ... 4? years ago, I think. Do you happen to know if they've improved an order of magnitude or two in the intervening years?
I feel you on the old-truck life support. My 99 F-150 is around at 250k, burning oil, and it doesn't make economic sense to swap or rebuild the motor. I'm going to hit up a GSA auction next week and see if I can score slightly 201x dinosaur-powered minivan instead.
(Score: 2) by JoeMerchant on Tuesday October 11 2022, @09:52PM (1 child)
Our last driver assistant rental was in 2019, the lane following could have driven for me, but knew if I took both hands off the wheel or stopped steering for myself even with hands on and would complain then disengage. It also did pretty good fixed distance following in traffic.
If your truck frame is in good (non rusty) condition, you might still be in "economically viable" territory for an engine rebuild or replacement... The rub comes in with automatic transmission control, fuel and spark control are pretty easy and cheap with aftermarket ECUs, but not so much with the different every year and model transmissions. The trick is: do you have a spare vehicle you can depend on for the weeks, possibly months a major drivetrain refresh would take.
Point of reference: I put a newly rebuilt and balanced 3 liter Jaguar V6 with bespoke adapter frame, flywheel, stainless exhaust and ECU iny 1991 Miata for $20k (and 3 months tweaking time) compared with a new Miata for $30k and up... If you are sticking with the same engine type you already have, it's much faster and cheaper.
Україна досі не є частиною Росії Слава Україні🌻 https://news.stanford.edu/2023/02/17/will-russia-ukraine-war-end
(Score: 2) by ElizabethGreene on Wednesday October 12 2022, @04:39AM
The frame is solid on it, but there are enough nit-pick things wrong with it that it I don't think it makes sense. e.g the headliner is gone, the upholstery failed and now the foam is gone under the driver seat, I've got some body rust from where Ford forgot how to Primer, etc. I can get a reman long block for $1200 and put it in myself, but the get-up-and-go of a new engine in it would probably shake something else loose in pretty short order. :)
To put it in "Hey, nice truck" shape is about $2500, which is about half what I'll spend on a new (to me) Grand Caravan. I'm partial to those; we've had one for most of the last two decades.
(Score: 2) by driverless on Tuesday October 11 2022, @07:52AM (2 children)
Yeah, except that a camera can't see through fog and rain and haze and oncoming headlights and other crap no matter how much Elon wants it to. He's basically nobbled his own product, hopefully he'll see the effect this footgun has and undo this brain fart fairly promptly. That is one thing he's got going for him, unlike politicians he doesn't continue down the wrong path in order to show resolve, he'll change direction if it's obviously the wrong one.
(Score: 2) by JoeMerchant on Tuesday October 11 2022, @10:22AM
Unfortunately for him, it feels like the social landscape isn't ready to make the changes necessary to make self driving cars (at anything over walking speed) a reality any time soon, whether they are guided by optical, ultrasonic, radar, lidar, or any combination of those at any price point.
I can't imagine that the net cost of components for ultrasound is even $100 per vehicle, including wiring, connectors and any dedicated processing hardware. There may be supply chain issues driving the change, at least in part.
Україна досі не є частиною Росії Слава Україні🌻 https://news.stanford.edu/2023/02/17/will-russia-ukraine-war-end
(Score: 3, Insightful) by RedGreen on Tuesday October 11 2022, @01:39PM
" That is one thing he's got going for him, unlike politicians he doesn't continue down the wrong path in order to show resolve, he'll change direction if it's obviously the wrong one."
Really so where does the Twitter decision fit into those stunningly great visions/decisions of his. He goes in and makes deal to buy it unconditionally then spends entire summer trying to get out of the deal talking trash about it, getting sued for trying to get out of the deal. Only to reverse course when it becomes obvious he will lose the lawsuit and be forced to pay for it anyways so now he is back to buying it. Hopefully he has poisoned the well good enough to kill off the piece of junk and lose all the cash, could not happen to nicer scumbag, his Aryan brothers whose speech he wants to enable without restriction, will be so proud of him for owing the libs...
"I modded down, down, down, and the flames went higher." -- Sven Olsen
(Score: 0) by Anonymous Coward on Tuesday October 11 2022, @03:17AM
> It feels to me like Tesla's in a death spiral.
Check their sales in China, no death spiral there:
https://www.bbc.com/news/business-62825830 [bbc.com]
> Tesla had its best ever month in China in September, according to its most recent figures, selling 83,135 cars.
(Score: 4, Interesting) by Runaway1956 on Tuesday October 11 2022, @02:08AM (9 children)
I've said it before: they need to ADD more sensors, of various types. Give the AI ten times as many senses as humans have. The more senses available, the more aware the damned AI can be. So what if they're redundant? Repetitive redundancy is GOOD if you're trusting your life to the system! Give it bat senses, give it radar, give it lidar, give it everything you've got, and dream up some more! Hey, ultraviolet light sensors could be good in some kind of situation. Raptors spot their prey with UV light, because rodent urine reflects UV. What do humans look like in UV? Can you spot the one who should have shaken a couple more drops off? That could be handy if you're trying to avoid running over humans.
Abortion is the number one killed of children in the United States.
(Score: 0) by Anonymous Coward on Tuesday October 11 2022, @03:15AM (1 child)
More specifically, Tesla needs demonstrate publicly that they are working on this problem:
https://www.advrider.com/advanced-driver-assistance-still-not-seeing-motorcycles/ [advrider.com]
(Score: 3, Funny) by Runaway1956 on Tuesday October 11 2022, @05:06AM
Exactly.
People have problems seeing motorcycles. It has always been blamed on a self-preservation circuit in the brain. People almost always see anything on the road that is big enough to be a threat to them, but they don't see motorcycles because bikes pose no real threat to the occupants of a car.
Musk's visual light sensors bring that self-perservation thing into question. The computer has difficulty seeing bikes, and the computer has no self-preservation circuitry. It fails to see small bikes, large bikes, fast bikes, bikes that are relatively almost stationary, yellow bikes, black bikes, bikes with little people on them, bikes with hulking goliaths on them, full dress bikes, and naked bikes.
There's a real problem there.
Abortion is the number one killed of children in the United States.
(Score: 2) by Opportunist on Tuesday October 11 2022, @06:11AM
I don't know if I want "too stupid to wipe his ass" to be a survival trait.
(Score: 3, Insightful) by istartedi on Tuesday October 11 2022, @08:07AM (2 children)
This reminds me of something. Humans use their eyes for almost all the sensory input when driving because they *must*. The eyes are augmented a bit with the feel of the road and yes even sound; but you could take those away and a human would still competently drive. For humans, it's all about the eyes so Tesla is thinking the AI should be able to work that way too. History suggests that while that's possible, it might not be optimal. What history?
I'm thinking about the history of flight. The Wright brothers used wing-warping, because birds use it. Birds use wing warping because when you're flesh and blood, it makes sense. An airplane isn't flesh and blood though. Flaps and hinges are much more practical to build in to a flying machine, and ultimately that's what everybody ended up using.
So while it's entirely possible that Tesla *can* make the AI drive the car with just two cameras, it's not entirely clear that it *should*. Multiple sensors as you advocate might ultimately turn out to be the more robust and practical way because a car is not an animal.
Appended to the end of comments you post. Max: 120 chars.
(Score: 3, Interesting) by JoeMerchant on Tuesday October 11 2022, @10:37AM (1 child)
The human two camera system also comes equipped with multiple advanced lens cleaning mechanisms, multiple ways to deal with problematic glare, bad sun angles, etc.
If we are going to rely on fixed mechanical cameras, I would go for about 30 of them to boost the MTBF up to something approaching what two human mounted eyes can deal with.
Україна досі не є частиною Росії Слава Україні🌻 https://news.stanford.edu/2023/02/17/will-russia-ukraine-war-end
(Score: 1, Touché) by Anonymous Coward on Tuesday October 11 2022, @06:49PM
> The human two camera system
As a human, I resent the comparison of my eyes to a dumb camera. My eyes have multiple layers of processing built directly into the back of the retina (sensor) and present my brain with much more information than a simple image.
(Score: 5, Interesting) by JoeMerchant on Tuesday October 11 2022, @10:31AM (2 children)
Sensor fusion is a messy problem, difficult to demonstrate absolutely consistent improvement through use of multiple sensor types, but very easy to demonstrate the edge cases where algorithms get confused and do dumb stuff because of the multiple input data types.
You are not wrong: more information can be better, but in a regulated life safety field with low risk tolerance, it's a very expensive and slow thing to develop.
Україна досі не є частиною Росії Слава Україні🌻 https://news.stanford.edu/2023/02/17/will-russia-ukraine-war-end
(Score: 3, Insightful) by FatPhil on Tuesday October 11 2022, @05:47PM (1 child)
Great minds discuss ideas; average minds discuss events; small minds discuss people; the smallest discuss themselves
(Score: 2) by JoeMerchant on Tuesday October 11 2022, @06:58PM
>Design it to require a supercomputer now, and it'll be running on your smartphone soon.
A couple of problems with that:
1) even if a supercomputer could help, it takes intelligence to make algorithms like "don't kill the motorcyclist" work. More often, when given a supercomputer the developers get lazy and use up the supercomputer power doing un-necessary stuff and still have problems making the algorithms meet the requirements.
2) cellphones are a screwed up ecosystem with outrageous barriers to entry and maintenance of the "apps" allowed thereon. Yes, they're powerful, but no: you can't expect them to reliably perform mission critical tasks, and that doesn't look likely to change in the coming decades.
Україна досі не є частиною Росії Слава Україні🌻 https://news.stanford.edu/2023/02/17/will-russia-ukraine-war-end
(Score: 3, Insightful) by fraxinus-tree on Tuesday October 11 2022, @07:40AM (2 children)
>No matter what car you drive, it's still not smart enough to get you from point A to point B all on its own.
Most professional drivers aren't much better either.
(Score: 0) by Anonymous Coward on Tuesday October 11 2022, @06:51PM (1 child)
Haven't been in a London taxi have you? To get that license you have to memorize the streets and pass a test. In my limited experience, there just are not any problems with navigation, with one of those drivers.
https://www.nytimes.com/2014/11/10/t-magazine/london-taxi-test-knowledge.html [nytimes.com]
(Score: 2) by fraxinus-tree on Thursday October 13 2022, @06:48AM
Well, London got it about right. Almost no one else did.
(Score: 1) by dwilson98052 on Tuesday October 11 2022, @04:23PM (1 child)
...will be easier without those pesky ultrasound sensors.
They've been one of the major reason why it's so difficult and expensive to get body work done any huge part of why Tesla has a history of disabling autopilot on cars that have body work done by unauthorized shops or individuals.
(Score: 2) by JoeMerchant on Tuesday October 11 2022, @05:43PM
My mother has a newer GM vehicle with the sensaround camera system, or whatever they call it.
Accidents still happen, and when she wiped out the front bumper cameras the repair bill ended up being more than double what it would have been without them - like up from $2000 to $5000.
Україна досі не є частиною Росії Слава Україні🌻 https://news.stanford.edu/2023/02/17/will-russia-ukraine-war-end