I agree. Level 3 is possible IMO. Level 4 seems to be happening within geofenced areas (Waymo). Level 5? That's a very, very tall order.So you are saying Tesla can reach Level 4 (no need for intervention) within the next ~5 to 6 years? My read of SAE and where we seem to be know tells me SAE Level 3 is a realistic (and likely what most people really mean when they say "Level 5" lol) goal in the next 5 years for the industry (Tesla, Waymo, or otherwise). Level 4 and Level 5 (as defined by SAE ) seem highly unlikely in the next decade IMO. BUT I'm no expert. Just a layman observer of this stuff so hopefully I am wrong
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
What if FSD doesn't materialize?
- Thread starter Goose66
- Start date
Regulations exist for many hazardous processes though, wouldn't you agree? If they have been written with due care and diligence, and approved by good practice that is generally acceptable. There is risk in everything, we know that. Anyone following the regulations is also protected so long as they have also exercised due care and diligence. The regulation board or the companies should not be worried unless they are negligent.
That's also why liability insurance exists, to cover you for mistakes. Of course, as you point out, death often gets people angrier than just having your poorly designed toaster catch fire.
FloridaJohn
Member
Which government agency (in the US) would be responsible for approving something like this? How does the application process work?
Knightshade
Well-Known Member
So you are saying Tesla can reach Level 4 (no need for intervention) within the next ~5 to 6 years? My read of SAE and where we seem to be know tells me SAE Level 3 is a realistic (and likely what most people really mean when they say "Level 5" lol) goal in the next 5 years for the industry (Tesla, Waymo, or otherwise). Level 4 and Level 5 (as defined by SAE ) seem highly unlikely in the next decade IMO. BUT I'm no expert. Just a layman observer of this stuff so hopefully I am wrong
"Can the car pull itself over safely if it finds itself outside it's operational domain?"
Yes vs No is the only practical difference between 3 and 4.
If a car can do 3, it almost certainly CAN do 4, technically. You just need to write your ODD specifically enough.
THAT said- going that route would be unlikely to meet what was promised to pre-3/19 buyers anyway at the least.
FWIW, Model 3 leases can't be renewed, nor do you even have the option to purchase at end of lease.
So anybody who bought the "old" version of FSD (when they explicitly promised at least L4 capabilities during the purchase) and leased a 3, at the end of the lease if that wasn't delivered, might have a case for getting their $ back at least (note- it'd only be a refund of the lease-pro-rated percentage of $3000 since this was before the big feature/price changes)
DanCar
Active Member
I'm thinking Tesla to get to Robotaxis in a meaningful way will have to add remote assistance. As a layman observer what do you see as the biggest obstacle(s)? Keep in mind that tech is improving rapidly. The field of neural networks and deep learning is improving constantly. Hardware is also improving. Even though tech is improving rapidly, we are no where near conscientiousness or true intelligence. It is just a bag of tricks. Some would argue a degree of intelligence is needed for robotaxis. I think remote assistance can handle the few cases where the software lacks confidence or true intelligence is required.So you are saying Tesla can reach Level 4 (no need for intervention) within the next ~5 to 6 years? My read of SAE and where we seem to be know tells me SAE Level 3 is a realistic (and likely what most people really mean when they say "Level 5" lol) goal in the next 5 years for the industry (Tesla, Waymo, or otherwise). Level 4 and Level 5 (as defined by SAE ) seem highly unlikely in the next decade IMO. BUT I'm no expert. Just a layman observer of this stuff so hopefully I am wrong
I'm hoping Elon isn't too stubborn with current design and to some degree admits failure and comes up with a new design. If you are thinking robotaxes won't work with current design, then yes that is possible. But I'm hoping with a more robust design:
1. more cameras
2. remote assistance
3. active learning: cars can learn from previous remote assistance
4. improved hardware: hw4, hw5, etc..
Then we will get to a robotaxi service or similarly: cars that you can sleep in.
Last edited:
DanCar
Active Member
Currently it is state by state. Waymo a few years ago said there interpretation of Arizona law, says it is legal. In other words there are no laws in Arizona prohibiting taxi service. So no approval required in Arizona a few years ago. Florida passed some legislation in the past year welcoming robotaxi service.
FloridaJohn
Member
Exactly. So we won't have to wait "years before regulators allow it," and at the current time, there is no "political liability for any regulator who approves something like this," since it is already "approved" in several states in the US. The only thing holding back Tesla from releasing autonomous driving cars is Tesla.
Tesla CEO Elon Musk drops his prediction of full autonomous driving from 3 years to just 2
Tesla CEO Elon Musk was at Stanford’s FutureFest recently and he described how he evaluates the progress and a timeline...
electrek.co
Elon Musk clarifies Tesla's plan for level 5 fully autonomous driving: 2 years away from sleeping in the car
Ever since Tesla released its second generation Autopilot hardware and started offering an option called “full self-driving capability”, there’s been...
electrek.co
Tesla Level 5: Autopilot
Recently Elon Musk publicly stated that it will have Level 5 autonomy by the year 2020. He went further to state that all Tesla’s will be…
towardsdatascience.com
Elon Musk brags that Tesla is "very close" to full, level 5 self driving
"I think there are no fundamental challenges remaining for level 5 autonomy."
futurism.com
FSD is just the latest iteration of level 5 autonomy from Elon/Tesla. I realize they went down that path initially with Mobileye and started off by offering Autopilot. But Elon has made many predictions about level 5 / full autonomous coming from Tesla - as far back as 2015. Now, to be fair, Google, Uber, and others have made claims and missed those timelines as well. So it's not like Elon and Tesla are out on an island promising things within certain timeframes that just don't materialize.
But Elon did throw a bit of a marker down with his "for sure" comments about >1m robotaxis by end of 2020. Even if you give them some pandemic leeway with that timeline, they still missed. I guess at this point, the question is do you believe Tesla is close to level 5 (despite nearly every other auto analyst disagreeing) or is this just another case of crying wolf.
Cheburashka
Active Member
Well, to be completely fair, none of those companies swindled money from consumers through promises for their fictional product but Tesla did.
Knightshade
Well-Known Member
Elon Musk on 60 minutes said:
Replace "mass-produced car" with "L5 autonomy system" and same idea applies.
He has no idea how long it'll take (and neither does anyone else since nobody has ever done it) but can't help himself throwing out dates anyway.
DanCar
Active Member
What is worse than throwing out dates is saying "for sure", "with certainty" and being completely wrong.
There’s of course a big difference between predicting something and selling it.
Knightshade
Well-Known Member
Sure.
Which is probably why in ~March 2019 they changed what they were selling to instead be a specific list of features.
Most of which already were available at the time, and of the remaining 2, one was delivered to customers last year, the last allegedly coming in the next month or two.
The "remaining 2" were both promised as "coming this year" in April of 2019.
Even if city streets navigation comes in June (LOL), that is 26 months on an 8 month written promise. And they took 13 months for stop lights when they also said it would be 8. Even as they have massively backed off on what they are selling as "FSD", they have still missed dates by large margins. They've missed their 2019 delivery of beta L2 city streets by longer than Elon estimated it would be until we had L5 back in 2016.
They haven't demonstrated a true level 3 yet which I believe is possible with the current hardware. Level 5 is out of the question judging from what's being shown. And I don't believe our cars currently possess the hardware or computing power necessary for a level 4 or 5 system. But, who knows, they might surprise us. I'm reading that he wants to do away with the radar input and rely just on vision. So we'll see how that goes.
IMO, level 3 is a more reasonable expectation which I'm OK with if they could pull it off.
Ruffles
Member
Is there any published agreed upon "conditions" for self driving? If you define the conditions, you can very easily reach various levels of automation. If you go the supercruise approach, you could say Tesla is very close to L4-5 on specific stretches of roads under specific weather conditions (i.e. no snow etc.). I suspect most people assume that an autonomous car would be able to drive in any condition they could but I don't think that is a fair expectation.
What if Tesla gets FSD working great but only in specific areas. Does that count? What if it works everywhere but only during the day? What if it works day & night but not in rain or snow? What if it can work any weather but only in certain places? There are all types of conditions that can come into play and manufactures can demonstrate success in different cases. Which conditions do you think the public will accept / expect?
For me, I trust and use AP on the freeways all the time. I keep my hands on the wheel and realize there are times (like in construction zones) that it may get it wrong. If they get the navigate on city streets to the same level as the current highways, I will consider it a success. I'm not looking to sleep in my car but I realize everyone has different expectations.
What if Tesla gets FSD working great but only in specific areas. Does that count? What if it works everywhere but only during the day? What if it works day & night but not in rain or snow? What if it can work any weather but only in certain places? There are all types of conditions that can come into play and manufactures can demonstrate success in different cases. Which conditions do you think the public will accept / expect?
For me, I trust and use AP on the freeways all the time. I keep my hands on the wheel and realize there are times (like in construction zones) that it may get it wrong. If they get the navigate on city streets to the same level as the current highways, I will consider it a success. I'm not looking to sleep in my car but I realize everyone has different expectations.
Knightshade
Well-Known Member
Read SAE J3016- it explains all the levels in detail.
The literal difference between L4 and L5 is L4 can have a limited set of conditions it works under, and L5 can't.
In BOTH cases though no human is ever required to operate the vehicle safely though. In L4s case, if it's going to leave it's set of conditions it must be capable of safely pulling over and parking itself without a human involved.
L3 differs from L4 in that it DOES require a human to actively take over if the car is leaving the set of conditions it can self-drive under.
As either L3 or L4 depending on the handoff requirement, yes.
Not L5. (Unless the "specific areas" is for legal reasons, not lack of capability)
I wonder why they have not released at least a Lv3 "trafic jam driver" up to 45 km/h max speed, 10 second driver standby, divided highway only? Don't they even trust their tech to do that? At least it could give some good marketing value. FSD as sold is something completely different..
DanCar
Active Member
FSD beta has to be released wider. For current autopilot / FSD, it has a number of issues, such as what happens when it hits a deep pothole or doesn't stop in time for an empty flatbed trailer.
Similar threads
