Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Is Tesla closer than we think?

This site may earn commission on affiliate links.
What’s silly in this thread is the mix of « what we know/see » and « what we guess ».

What we know as a fact is that Waymo has L4 with about 100k per disengagement. Making fun of the cone incident? Sure, but no real need as this will be counted against the 100k disengagement statistic, so focus on the number, not on the trolling comment, if you care about tracking progress.

What we know as a fact is that Tesla has L2 with a much lower disengagement rate and that they are not doing L4 testing in CA,.

That’s what we know.

Then, the rest are just guesses and opinions: L2 being better than L4 - or not, Tesla doing L4 outside of CA, Waymo being able or not able to operate in any city in the USA if they wanted to, etc. And anybody thinking that Elon Musk saying something means it is true, I’d suggest you join the thread of whether a class-action on FSD is possible as half the running opinion here is that whatever Elon says is just noise at best.
 
On « Tesla can do L2 anywhere unlike Waymo », I just want to add that you might want to test a Tesla in different countries before saying so. My tesla breaks when seeing traffic lights… for train tracks next to the road, can’t handle any roundabout which we have 10x more (at least) in Europe vs. stop signs, doesn’t handle merge on most semi-highway, mis-read a bunch of speed signals and a lot more situations that I’d call relatively basic. So, yeah, more training with geo-specific pics, and regulations, etc. But my point is to split what’s factually true from what’s opinion. TODAY, a Tesla doesn’t operate as smoothly in all countries, that’s a fact, so beware when making comments about Tesla being much more general purpose than Waymo, that’s an opinion, you don’t KNOW for sure.

Also, a good point was raised above: the legal responsibility associated by being L2 or L4 is vastly different, so drawing 1:1 analogies between what Waymo and Tesla are doing similarly or not is also a risky slope. We essentially know that Tesla is taking very very minimal risk by deploying their L2 anywhere in the world « because you have to always be in control » (and anytime a problem occur, we are reminded on this forum that the driver was a moron not to be more cautious). But with L4, if I was the person in charge of the rollout, I’d probably also add cities step-by-step, adding increased complexity and use cases over time, until I get to 1m miles per disengagement. Expanding to « all of the USA » doesn’t necessarily make sense if the maturity improvements they have to work on can be all faced by expanding the scope gradually, making it logistically easier. Yes, maybe it just has to do with logistics. Again, my point is that: we don’t know what we don’t know, so any argument saying that Waymo is inherently geo restricted as a L4 is pure speculation (much like the opposite).
 
Well my post was throwing your logic back at you to show you how flawed your logic is. I guess you missed it.

What we know is that Tesla has deployed L2 everywhere. Waymo has deployed L4 but only in limited geofenced areas so far. They are testing in other areas but it is not reliable enough yet. That's understandable since Waymo has not solved FSD yet. Nobody has.

Tesla can deploy L2 everywhere because L2 is a lower bar. L2 does not have to work with no human intervention Tesla can deploy L2 that is still beta because the driver is expected to supervise. Waymo does not do L2. Waymo only does L4. This means that Waymo could have FSD that works everywhere but they will only deploy it to the public when it reliable and safe enough to be driverless.
Minor quibble but my understanding is that level 4 by definition is geofenced.
 
Minor quibble but my understanding is that level 4 by definition is geofenced.
L4 Could be geofenced, or it could be limited in other aspects that would be specified in the ODD. In Waymo's current case, it's geography, but I would imagine as more systems get deployed, you will see it limited by things like adverse weather, etc. Whereas level 5's ODD is all human driveable conditions.
 
I love the discussion, but it seems that I'm the only that believes my original hypothesis could happen.

Let's for a second then assume that there is a huge limitation by the current sensor suite. I like to use @Chazman92 Left turn over 3 lanes as an example.
Sor for this example, we will say that they find they cannot get to 99.999999% safety crossing more than 2 lanes of traffic doing a left turn, but in every other situation they can meet the safety requirements to deploy. Do you guys really think they wouldn't just adapt to avoid those situations??

 
I love the discussion, but it seems that I'm the only that believes my original hypothesis could happen.

Let's for a second then assume that there is a huge limitation by the current sensor suite. I like to use @Chazman92 Left turn over 3 lanes as an example.
Sor for this example, we will say that they find they cannot get to 99.999999% safety crossing more than 2 lanes of traffic doing a left turn, but in every other situation they can meet the safety requirements to deploy. Do you guys really think they wouldn't just adapt to avoid those situations??


Yes, it is definitely possible and it makes perfect sense. I just think Elon is stubborn. Imposing an ODD limit like you suggest, would make it L4 at best an Elon insists that the hardware is capable of L5. So Elon will not impose limits that would imply he was wrong. He will just keep driver supervision and insist that they are training the NN to solve those problems, and claim some new rewrite will solve the problems in 3...6 months. That's my thought.
 
Minor quibble but my understanding is that level 4 by definition is geofenced.
L4 Could be geofenced, or it could be limited in other aspects that would be specified in the ODD. In Waymo's current case, it's geography, but I would imagine as more systems get deployed, you will see it limited by things like adverse weather, etc. Whereas level 5's ODD is all human driveable conditions.

Booman,

Microterf is 100% correct. L4 can have a lot of different constraints on the ODD. Geofencing is the most common but the constraint could be weather, speed, traffic conditions, road type, road conditions, time of day, etc...

It is very important to note that L4 constraints are NOT situations that the autonomous vehicle is not good at. The SAE levels are NOT performance based. So an autonomous vehicle missing a stop light or getting confused by a construction zone does not make it L4.

L4 constraints on the ODD are "artificial" constraints that the manufacturer places on the autonomous driving system that limit the ODD to something less than what you might expect from a "typical skilled human driver" (SAE term):

Examples of L4 constraints on the ODD:
- An autonomous vehicle that always reroutes to avoid highways
- An autonomous vehicle that only operates on highways
- An autonomous vehicle that always reroutes to avoid construction zones
- An autonomous vehicle that always reroutes to avoid school zones
- An autonomous vehicle that only operates inside a geofenced area
- An autonomous vehicle that always drives at less than 35 mph
- An autonomous vehicle that only operates between 7am and 7pm
- An autonomous vehicle that always stops if it rains

etc...

In contrast to L4, we can think of L5 as autonomous driving with the same ODD as a "typical skilled human driver". So L5 might pull over in a severe thunderstorm for safety like a human might but the L5 would be free to drive around much like a human driver would. So for example, L5 would drive on any accessible public road, not be geofenced, drive any time, day or night, drive any speeds up to speed limits etc... like a human driver might.
 
Last edited:
I love the discussion, but it seems that I'm the only that believes my original hypothesis could happen.

Oh, no, I think in some form it is almost guaranteed to happen, since Karpathy described this as part of the plan in his most recent public talk. There is a lot left unknown about how this will be implemented though, and how much of a difference it will make in pushing Tesla to a greater degree of autonomy.
 
I’m basing it on 5 years of ownership experience and the recurring problems that spring anew.
”Yikes, Truck Lust!” becomes “Argh, Phantom Braking!” becomes <insert 2022 problem here>.
well, there are always new problems to deal with till your dead... that’s how life goes.
I was just looking at the timeline and remembering what was delivered in those 5 years that @BrerBear is complaining about.

I mean, Navigate on Autopilot (with no confirmation lane changes), Stop Sign and Light control are some big ones.
But, just going over a hill (cresting) and to finally not have Autopilot freak out.

People forget really quickly just how many improvements there have been in the 5 years.
 
Also summon. No one else comes close to Tesla in features and usability.

People, especially the eternal complainers, aren't forgetting anything. They're reduced to whining about autosteer city because it's the only feature that hasn't been released widely.

I was just looking at the timeline and remembering what was delivered in those 5 years that @BrerBear is complaining about.

I mean, Navigate on Autopilot (with no confirmation lane changes), Stop Sign and Light control are some big ones.
But, just going over a hill (cresting) and to finally not have Autopilot freak out.

People forget really quickly just how many improvements there have been in the 5 years.
 
  • Like
Reactions: mikes_fsd