Autonomous Car Progress

[eli_]

Yes - I understand all about having a sub-par standard is better than no standard at all.

But some "standards" are so poor, they should be thrown out. Levels are one of those. What were they thinking ... or more accurately why were they not thinking ? I understand compromises, behind the door politicking that is inherent in all standard setting. But this one is just impractical and more importantly dangerous - as we see with the Honda L3 Legend.

This kind of standard should give a clear idea about the ability (which includes quality) of systems. To completely "forget" about quality is inexcusable. So, I refuse to have them dictate our conversations.

SAE standard levels get an "F" from me.

SAE levels just boil down to who is liable for accidents IMO. I.e. you can have the most advanced driving robot in the world (99.9999% reliability on every road anywhere) and you're still classified as level 2 because the driver must be present to avoid the OEM taking liability (why would they ever do that?). And the reverse is also true, you can be a level 4/5 glorified streetcar that only drives on a fixed route but doesn't require a driver because the taxi company owns the cars and takes liability.

 

[Daniel in SD]

LOL.

What is really crazy are the assumptions he makes. Lets look at this one in particular.


View attachment 769036

Yes - thats right, Brad is so out of touch with the real world he thinks "out of date maps are rare".

You can see the TomTom maps thread I opened to talk about some of the mapping issues I found not far from my home in my sig.

Let me just explain this
- I do not live in the sticks but in a city just miles from Microsoft HQ. Its a city that is within top 5 when it comes to per capita income in WA.
- When I started having peculiar issues with FSD I (with the help of several on this board) started investigating and figured out that maps were outdated
- So how many issues did I find ? More than 20 issues within 2 miles of my home !! Not just that - these are roads I take regularly and would have never known had issues but for FSD. These are not roads I went out of way to find issues in - just regular roads near my home. Infact atleast half the roads I regularly use have some kind of problem.
- Just yesterday I went on a road I don't use often and found 4 round abouts that were missing in the map
- Some parts of the map have not been updated for over a decade when my neighborhood was built !!!!
- I've got more than 2 dozen updates done to TomTom maps near my home.
- Ironically, Bing uses TomTom maps. Half the Bing team lives my my city

I guess, Brad's assumption shows what happens when you live in the self-driving industry capital of the world. The maps near your home are always updated quickly ! Never mind 99.999999% of rest of the world.

When I worked in Microsoft we used to have a map that showed half the world covered by Redmond, WA. As we went further geographically the area became small. Huge countries were just a speck of the dust. Showed how easy it was to be "Redmond" centric. Brad (and other AV industry vets) need that kind of a map about the silicon valley.

"Out of date maps are rare" .... my ***.

If only Tesla had a fleet of vehicles that they could use to detect map errors like AV companies do.

 

[EVNow]

If only Tesla had a fleet of vehicles that they could use to detect map errors like AV companies do.

Right.

Instead of this weird obsession some have about HD maps, Tesla should be concentrating on getting their SD maps updated quickly. There should be automation as well as simple way for us to send updates.

Apparently TomTom gets 10k manual map update requests a day and they handle it efficiently. Why can’t Tesla do that.

 

[Daniel in SD]

why would they ever do that?

Robotaxis! It's weird to hear other Tesla fans downplaying the value of driverless vehicles.

https://twitter.com/x/status/1450868052040507392

 

[Daniel in SD]

Right.

Instead of this weird obsession some have about HD maps, Tesla should be concentrating on getting their SD maps updated quickly. There should be automation as well as simple way for us to send updates.

Apparently TomTom gets 10k manual map update requests a day and they handle it efficiently. Why can’t Tesla do that.

Maybe they're working on automating generation of HD AV maps and don't want to waste time on SD maps.

 

[eli_]

Robotaxis! It's weird to hear other Tesla fans downplaying the value of driverless vehicles.

https://twitter.com/x/status/1450868052040507392

Yeah I haven't bought into robotaxis being valuable yet, and I don't think any OEM would take liability for accidents in a personal vehicle that they don't own or maintain. If it's fleet owned, then sure. Otherwise it's total nonsense, if you're doing a driver assist thing like FSD I would be Level 2 and proud, be level 2 forever.

 

[EVNow]

Maybe they're working on automating generation of HD AV maps and don't want to waste time on SD maps.

My bet is - having talked so much about HD maps, Elon doesn’t want to bring up the reliance on (any) maps. He is probably stubbornly sticking to “but humans can drive without maps”.

 

[EVNow]

SAE levels just boil down to who is liable for accidents IMO. I.e. you can have the most advanced driving robot in the world (99.9999% reliability on every road anywhere) and you're still classified as level 2 because the driver must be present to avoid the OEM taking liability (why would they ever do that?). And the reverse is also true, you can be a level 4/5 glorified streetcar that only drives on a fixed route but doesn't require a driver because the taxi company owns the cars and takes liability.

What will drive OEM taking responsibility is market pressure.

If Apple car and Tesla are equally good at FSD and Apple offers to take responsibility, Tesla will have to, too.

 

[Knightshade]

Robotaxis! It's weird to hear other Tesla fans downplaying the value of driverless vehicles.

FWIW I believe robotaxis will be incredibly valuable.

Like add a 0 to your market cap even if you're already a trillion dollar company valuable.

What I don't believe it that Tesla will have any by the end of this year. In fact I'd be shocked if they did.

 

[eli_]

My bet is - having talked so much about HD maps, Elon doesn’t want to bring up the reliance on (any) maps. He is probably stubbornly sticking to “but humans can drive without maps”.

It just comes down to the basic product requirements, Tesla makes a driver assist product (no evidence of a robotaxi) so it has to work everywhere, in every random owner's neighborhood street or dirt road, or it would be a pretty crappy product. And that means the intelligence has to be entirely on the car itself. They can't "drive on maps" like Waymo can, the static problem of driving (road geometry, lanes, signs, etc.) has to be solved in the car in real-time. For Tesla, maps are for "tips and tricks", i.e. semantics about the lane the car is in (there will be a turn restriction in ~1.5 miles I should make a lane change now). It's better if they have accurate maps but the moment to moment driving has to work offline using vision only.

 

[bradtem]

Can you expand on this ? What kind of semantic information that everyone puts and what kind of info ME is not putting it ?

To me the interest is mainly in consumer AVs and three players seem to be serious here. Tesla, ME and GM. It would be interesting to see how each of them is handling various parts.


What Tesla does with maps is bit of a mystery. A lot of FSD problems can be traced to simple mapping issues but there is no way to get them quickly corrected. Tesla needs to be more transparent about this.

ME has a mission of shrinking down their map info to 10kb/km, which is probably too far. So they try to isolate only what is most important. Others put in more because they don't have that limited a data budget.
GM is mainly focused on robotaxi. Even MobilEye is first on robotaxi because it's much easier than a consumer car so it's the obvious first step for many teams. Tesla wants to do robotaxi, but second.

Semantic information includes special rules and meaning about any lanes and locations, road rules and meaning of signs. Meaning of different traffic signals. Typical driving patterns in the lane, where drivers go and don't go, where they stop and don't stop, where they slow, where they move to in order to see cross traffic, restrictions on lane use, good places to go in an emergency, where pedestrians tend to cross and what they do. Whatever would be useful.

 

[Dewg]

Fun video to watch on FSD Beta and recent attacks against it:

 

[diplomat33]

https://twitter.com/x/status/1493614703007969282

 

[EVNow]

FWIW I believe robotaxis will be incredibly valuable.

Like add a 0 to your market cap even if you're already a trillion dollar company valuable.

What I don't believe it that Tesla will have any by the end of this year. In fact I'd be shocked if they did.

I think the problem with evaluating value addition with robotaxi service is - the revenue & margin are dependent on when FSD is achieved and how many other players are there at that time. Then, there is the pesky issue of getting permit in each city separately. If Tesla gets to FSD next year vs in 2030, the money they can make is different.

What is for sure is - each Tesla car will be more valuable with city FSD feature. They can probably charge a hefty subscription fee and have a steady stream of revenue from it.

 

[Knightshade]

I think the problem with evaluating value addition with robotaxi service is - the revenue & margin are dependent on when FSD is achieved and how many other players are there at that time.

I'm referring to whomever gets there generally, and first. Which Elon at least insists will be Tesla.

If Tesla arrives 8th, and years after the Mobileyes of the world have been operating robotaxis for 5 years I expect the impact to be pretty insignificant in comparison.

Then, there is the pesky issue of getting permit in each city separately

Which is a nonsense red herring repeatedly debunked.

It's already legal in half a dozen entire US states.

some of them even specifically cite ride hailing/fleet use as explicitly permitted today- and nearly all of them forbid smaller governments (like cities) from making their own separate rules on it.

Once someone actually has a working system to roll out they can do it immediately in those states, no "waiting" for anyone or any thing.

Once they do, and it proves safe and effective, many other states will follow along fairly (for government) rapidly.


Places like the EU might take a bit longer to figure out regulations- but public transit is already worlds better there than here so it'll be less of a revolution anyway.

 

[EVNow]

Tom Tom's map is not a AV HD map...
You are actually proving the AV industry right about HD maps. You are proving that using some off-the shelf map and combining it with your internal map with undisclosed level of detail just to avoid "using" HD map is a hilariously horrific approach.

I'm proving you guys have no idea about real-world where even SD maps have a lot of holes and are outdated - let alone HD maps.

All you are saying is HD maps will be "magically" always updated and much better and cheaper than SD maps. And you actually don't need the maps to be accurate because you can drive without accurate HD maps - but you MUST have HD maps. Looks more like a pretzel than a rational argument.

 

[EVNow]

It's already legal in half a dozen entire US states.

some of them even specifically cite ride hailing/fleet use as explicitly permitted today- and nearly all of them forbid smaller governments (like cities) from making their own separate rules on it.

You mean red states like Mississippi ? Blue states where most of the robotaxi money is usually have city control.

ps : Yes, I should add an addendum to my original statement. You may be able to buy state legislature to circumvent cities - but in cities with the most robotaxi money it might be difficult. Try getting NY state to overrule NY city.

 

[EVNow]

GM is mainly focused on robotaxi.

You mean Cruise - whose CEO was recently fired ? Have you seen their "UltraCruise" plans ?

GM’s Barra Dismissed Cruise CEO Ammann Over Mission, IPO

Cruise LLC Chief Executive Officer Dan Ammann had a slate of meetings on Dec. 16 when he got an early afternoon call from General Motors Co. CEO Mary Barra. She told Ammann he was being dismissed from the robotaxi startup that GM controls through a majority stake, say people familiar with the...

www.bloomberg.com

Cruise CEO Dan Ammann Fired By GM's Mary Barra Amid Differences Over IPO Plans, Company Focus: Report - General Motors (NYSE:GM)

General Motors Co (NYSE: GM) CEO Mary Barra has dismissed self-driving subsidiary Cruise’s CEO Dan Ammann due to differences over the focus on the robotaxi unit, Bloomberg News

www.benzinga.com

Barra and Ammann reportedly had differences over the focus of the robotaxi business, with the GM board being inclined towards expanding the self-driving unit to create luxury Cadillacs and self-driven cars that could be sold at retail stores.

Ammann was open to ideas but wanted the company to focus on the robotaxi unit before diverting the resources towards expansion, as per Bloomberg.

Regarding ME - this is not what the CEO talked about at CES. Infact he talks about the pitfalls and inability to scale as the problems of robotaxi focussed AVs. It seems he prefers Consumer AVs.

Even MobilEye is first on robotaxi because it's much easier than a consumer car so it's the obvious first step for many teams.

 

[Knightshade]

You mean red states like Mississippi ? Blue states where most of the robotaxi money is usually have city control.

Well, no, I don't mean Mississippi since that's not a state it's legal in.


Granted, SOME of the state it IS legal are red..... Texas, North Carolina, Florida, Tennessee, and Nebraska.

Though not all- Georgia and Colorado were both blue last election and allow self driving cars today too.

That actually covers half of the 14 most populous cities in the US BTW and 11 of the top 25 per this list:

The 200 Largest Cities in the United States by Population 2024

Which is pretty good given it's only like 7 out of 50 states.


And in some others (Illinois for example) they haven't explicitly allowed unregulated self driving cars but they HAVE passed laws prohibiting individual cities from regulating them.




so no, "waiting for regulators" isn't what's going to stop robotaxis from establishing if they're gonna be a thing or not.

They can be rolled out today in a bunch of big cities if someone wishes.

Plus, without needing a human, it can vastly expand where it makes any sense for there to BE a taxi market.

It makes no sense today to have taxis in my area for example. The human needed to sit and wait for the rare call would kill the economics of it.

But an RT that can just sit there basically sleeping for 0 marginal cost until someone in my rural area needs a ride? Suddenly a market that didn't exist before does. It's not gonna match airport runs in NYC, but you can repeat it in a MASSIVE # of rural areas with 0 service today.

 

[bradtem]

You mean Cruise - whose CEO was recently fired ? Have you seen their "UltraCruise" plans ?

GM’s Barra Dismissed Cruise CEO Ammann Over Mission, IPO

Cruise LLC Chief Executive Officer Dan Ammann had a slate of meetings on Dec. 16 when he got an early afternoon call from General Motors Co. CEO Mary Barra. She told Ammann he was being dismissed from the robotaxi startup that GM controls through a majority stake, say people familiar with the...

www.bloomberg.com

Cruise CEO Dan Ammann Fired By GM's Mary Barra Amid Differences Over IPO Plans, Company Focus: Report - General Motors (NYSE:GM)

General Motors Co (NYSE: GM) CEO Mary Barra has dismissed self-driving subsidiary Cruise’s CEO Dan Ammann due to differences over the focus on the robotaxi unit, Bloomberg News

www.benzinga.com

Regarding ME - this is not what the CEO talked about at CES. Infact he talks about the pitfalls and inability to scale as the problems of robotaxi focussed AVs. It seems he prefers Consumer AVs.



View attachment 769431

Yup, that's the Cruise I mean, and the firing of the CEO that I wrote an article critical of. Yes, he wanted to do all robotaxi, and Barra wanted to take more tech and put it in GM cars, but that doesn't mean Cruise is not itself primarily working on Robotaxi, just that they don't get to go 100% on it as wanted.