Yes, but it appears it's the only item that people are obsessing over right now. Many here equate FSD with Autosteer on City Streets.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Analysis of the price-hike for FSD, and the options it allows Tesla
- Thread starter bradtem
- Start date
Tam
Well-Known Member
2022 Mercedes-Benz EQS L3 Drive Pilot has already been sold in the EU for about 7,430 euros ($7,726). Awaiting for sale permit in California.
Come on... Admit it
Here's the proof in the picture included in the article: The FSD beta Blue Path Depictor shows that it's in the LEFT turn-only lane, while the routing map shows that it needs to turn RIGHT!
This is why Tesla removed the radar to reduce the confusion about sensor fusion. Now, there are still 2 confusing decisions, what should be removed next?
Anyway, Tesla is trolling us because they should have fixed this kind of mistake first before raising the price to $15,000.
Photo: Forbes
Funny, I drive through this intersection as part of my FSD test route every update. It will eventually get into the right lane crossing over the solid white, but as you said, this is definitely incorrect. Usually I’ll manually execute a lane change to the right lane before this comes up. But up till 10.12.2, this hasn’t been fixed
Ultimately it comes down to three things.
How Tesla treats customers who were promised more than Autosteer on City Streets? Can they simply do a partial refund to satisfy them?
Do regulatory agencies like the NHTSA put a stop to FSD because of obvious safety concerns when a car does ALL the driving, but expects the driver to intervene even when it works >99% of the time.
How Tesla treats all HW3 owners when HW4 comes out. If they update owners to the latest and greatest I think most customers will be satisfied, but if they don't they're going to be pretty livid.
Tesla has to do some kind of creative dance when they really did promise customers things like Robotaxis, but only delivered an L2 system that can't even compete on self-parking.
Does anybody have the subscription terms? I presume you don't get the computer upgrade, and if you have it, I really doubt they will do any hardware upgrade that's needed later.
AlanSubie4Life
Efficiency Obsessed Member
It’s a $1k upgrade, for subscriptions. For now.
Looking forward to hearing about HW4 transition very soon! They have definitely been keeping that hush-hush but should be a trivial retrofit. Though expensive. All very hard to say how it would work. Obviously not clear whether it will happen.
Full Self-Driving Computer Installations | Tesla Support
While no Tesla vehicles are fully autonomous today and require active driver supervision, the FSD computer is capable of delivering intelligent performance and control to enable a new level of safety and autonomy, without impacting cost or range. Learn more about FSD computer installation.
More importantly, and what everyone is dancing around, is whether Elon's Twitter comments are legally actionable (promissory estoppel). No where on Telsa's site can you find anything about your car becoming a robotaxi. It was never mentioned there, nor mentioned when people purchased their cars.
Tesla does say that HW upgrades are "complimentary" in their language regarding FSD:
"If you purchased Full Self-Driving capability and have Autopilot computer 2.0 or 2.5, you are eligible for a complimentary upgrade to FSD computer. Certain early production vehicles may also receive complimentary camera replacements. A complimentary hardware upgrade is not available for Full Self-Driving capability subscriptions; however, you may be eligible to upgrade at your cost." (source: Full Self-Driving Computer Installations)
You'll notice they were a little cagy with "certain early production vehicles" - perhaps giving themselves an out.
Regulatory agencies are in their infancy with regards to FSD from all vendors. Laws and regulations are mostly not written yet, and the agencies are cognizant of their power to help or destroy a fledgling industry, so they are proceeding slowly.
We're living in a very interesting time.
Tam
Well-Known Member
I have no training in technology, so my thinking comes from the internet.
1) I don't think FSD will ever work with the current hardware and software philosophy. Its cameras can't even do a birds-eye view. The thinking is that since a human has 2 eyes and Tesla has 8 cameras, its system should be superhuman. That's the same thinking that if a bird has 2 wings and my car has 8 wings, my car can flap and fly better than birds. I see ornithopters in the movie Dune in the year 10191 so someday that kind of thinking will be realized but not during my lifetime.
2) MobilEye says that Vision is good for L2, but beyond that, you need LIDAR or 4-D RADAR.
3) Waymo has proven that L4 has worked with no fatalities since its first public rides with no Waymo staff or drivers in 2020 for the past 2 years. The remaining issue is when a consumer can own one.
4) I think it's possible to have a very good L2 with seldom a fatality/collision by using Waymo technology. I don't think I will see L5 (Tesla's way of making money picking up rides while the driver sleeps at home) in my life from any company.
Additional infrastructure might be needed: V2X (Vehicle-to-everything: from signal lights to other cars to crossing guards...)
5) I don't think we'll get L5 during our lifetime. However, it could be watered down to a safe autonomous vehicle that doesn't cause a collision but might get stuck in some situations requiring a driver or a tow.
6) I don't think Tesla will ever get L3 due to:
a) Technicality: MobilEye explanation in #2 above: lacks LIDAR or 4D-RADAR.
b) Legality: EU law mandates that manufacturers are responsible for any accidents during the operation of L3. It's doubtful Tesla will accept such a law.
Last edited:
diplomat33
Average guy who loves autonomous vehicles
I would add that one big difference between Mobileye's vision-only and Tesla's vision-only is that Mobileye uses crowdsourced maps and a proprietary driving policy. I think this makes a big difference. I think Mobileye's maps help it avoid a lot of the map issues that Tesla's FSD Beta seems to have. Also, Shashua mentioned that the maps allow the car to see beyond visual range which helps the car plan better. And Mobileye's RSS driving policy helps it handle maneuvers like cut-ins and yielding safely. I firmly believe that FSD Beta would be much better, likely even hands-free L2 if Tesla also used crowdsourced maps and a robust driving policy.
On a side note, I have no problem with vision-only L2. Clearly, vision-only can do self-driving with driver supervision, especially when it is augmented with maps and a good driving policy. I think that radar and lidar are needed for L4 because to remove driver supervision entirely, you need more reliability and more redundancy than just vision. That's why I really like that Mobileye is pursuing vision-only for L2 but adds a radar-lidar stack for L4.
Last edited:
AlanSubie4Life
Efficiency Obsessed Member
From which manufacturer? I am not aware of any good driving yet. Maybe in very very limited ODDs (wide open freeway in daylight with no other traffic seems quite good).
In general I am not aware of any good L2 driving regardless of sensors but I don’t pay too close attention to Mobileye so maybe they are awesome right now?
Last edited:
diplomat33
Average guy who loves autonomous vehicles
I removed "good" since that is subjective. But vision-only can do self-driving with supervision. Both Tesla and Mobileye have self-driving with supervision now. And it stands to reason that vision-only will get better, not enough for L4, but probably "good enough" for L2 "FSD".
Total aside, apologies - but I looked up RSS and started reading the material on their website. I nearly did a spit-take when I read this:
-------------------
Don’t hit the car in front of you
Based on common laws of physics and a century of driving experience, new human drivers are taught to leave enough space between themselves and the car in front of them in order to provide enough time to react if the car in front were to suddenly brake.-------------------
Clearly they haven't driven in Los Angeles - LOL!! I mean they're right - but good Lord people in SoCal don't know how to drive...
Regarding the HW3 -> HW4 upgrade path, I feel like that question has already been semi-answered when they charged HW2.5 owners $1000 to upgrade to HW3 after we were sold cars that had "all the hardware necessary for full self driving capability". I wonder if anyone has got that upgrade installed for free through arbitration? I'm guessing if they did, they're under an NDA clause. Can anyone here say that they submitted for it and were rejected?
When did they charge HW2.5 owners $1K for HW3? Do you mean owners who bought FSD? If they don't buy FSD, do they need HW3? My impression was if you bought FSD and had HW2.5 they upgraded it for free. That's what they did for me and I paid only $2K for FSD.
diplomat33
Average guy who loves autonomous vehicles
We cannot make any definitive judgements on total safety or reliability based on just curated videos but what we see from Mobileye so far looks good IMO. It is especially impressive that the same vision-only system seems to work well in very diverse situations. Mobileye is testing in 10 different cities around the world. Additionally, Mobileye says they completed a 2000 km "FSD" drive across Europe with only occasional interventions.
Here are some examples:
New York
Tokyo
Munich
Jerusalem
AlanSubie4Life
Efficiency Obsessed Member
This is one of those silly arguments around here where people who did not buy FSD are upset that they have to pay for HW3 if they go with a subscription model. You can review other posts for the entire sordid history of people arguing back and forth about it.
AlanSubie4Life
Efficiency Obsessed Member
I did not have to look very far.
Go to 6:30 in the first video to see the car blatantly miss a green light. This is not a good L2 system; I would blast through this with accelerator application, being sure that no pedestrians are tempted to enter the intersection.
It may be just fine for an L3 or L4 system, but not helpful as an L2 driver assist (it needs to be quite a bit better in certain ways!).
Also no audio so you cannot hear all the honking, lol. (Note the human driver blasting around the ego vehicle in the above situation, since it was being ridiculous.)
This is a really tough problem and I encourage people to actually use the systems in question before determining whether they are good.
Last edited:
My standard that I tell the companies is that while demo videos are nice and a starting point, what you really need are demos where an outsider (customer, member of the press) picked the time and the route. To give Tesla full credit, because they let customers try it out, they do this the best. The only problem, of course, is that the results suck. Maybe everybody else's results also suck, though Waymo, Cruise, AutoX, Baidu, WeDrive and a few others also allow customers to order rides even with safety driver on board in some cases. And a few more companies like Motional and MobilEye will also soon be doing this. Tesla is the only one that lets a customer try it anywhere in the USA, which is amazing -- again, if the result did not end up sucking it would get full marks.
Similar threads
