I’m all for the development of robo taxis. However, I think most are over estimating the cost advantage of removing the driver who partially just gets replaced with cleaning crews and remote operators.
Let’s go through a short list items…
A driver can quickly decide if the car needs serious cleaning or not in a second, can prevent most serious cleanups just by being there, can ask the rider is a slightly different drop off location is OK when there is a lot of congestion and never has to call home for help unless in a serious accident. The driver/owner also takes care of all minor maintenance, never really needs to clean cameras but can clean windows as needed and usually has their own paid-for parking space. They can help elders and others with heavy luggage.
Of course drivers can be annoying and chatty, get tired and be creepy to single female riders.
And most drivers do this as an as-needed side hustle so they don’t consider time off, vacation and retirement as part of what they signed up for.
Taxis aren’t racing on a track, but perhaps the larger question is how something like LiDAR detects smell or feel or even sound. Because no-one here seems to think that Waymo will have problems due to its lack of smell detection.
As for “feel” Teslas already have torque sensors in its steering rack and in velocity detectors in its shocks, as well as those used in its airbags.
From your examples, you’re obviously aware that Waymo does this. Tesla owners know that for $99/year they get HD video streaming services to their touchscreen. I’m not sure what hurdle you think is present here. Tesla isn’t paying individual retail consumer rates for bandwidth.
As Tesla expands its service area, the issue of cellular reception quality will matter, as it will for Waymo and all other players. Note that Tesla’s FSD already operates exactly the same in areas without any cellular reception (I’ve tested it in remote areas) - the connectivity issue comes up with regards to issues that come up, like the car asking for a path in some unique construction situation. OTOH, I wouldn’t be surprised to see Tesla robotaxis that can venture outside populated areas to have Starlink built-in.
Netscape didn’t have a sustainable business model, as Microsoft simply packaged their web browser for free within its products. Who’s going to do something similar with robotaxis?
OTOH, Tesla has a unique business model, in that they can not only run a taxi service themselves, they can (they say) let Tesla vehicle owners put their cars into service to earn money for both Tesla and them, plus there’s a whole business around selling vehicles that can drive their owners, friends and families around.
It’s fair to question whether Tesla can achieve those goals, and within what time frame, but the comparison to a software product that became a feeebie in the world’s most popular operating system isn’t appropriate.
It’s not so much the competition, as Musk not keeping his eye on the ball. Musk has shifted Tesla from caring about the EV market to caring about autonomy and robotics. If he cared about the EV market, he wouldn’t have cancelled the $25k Model 2, for instance. So, self-inflicted. And yeah, the danger here is if Musk decides that robotics is more important than vehicle autonomy and shifts Tesla’s emphasis that way prematurely.
This has always been the argument against TSLA. Before 2013, the probability of Tesla just surviving as a company was in question, especially since the US hadn’t had a new car company survive in several decades. That changed with Model S. Then the question as to how popular Tesla could make EVs was in question. That changed with Model 3 in 2018 or so and then Model Y blew the doors off the stock by 2021. In retrospect, those don’t seem like they were “low probability” events, but many at the time considered them so.
And then Musk bought Twitter, lost his focus on Tesla, gave up on just plain ole’ EVs for autonomy, and then got heavily involved in politics - and the stock reacted appropriately.
Today, with the stock in the $320s, it would appear that Mr. Market doesn’t share your view that autonomy and robotics are “very low probability” events. Are they less probable than Tesla’s past achievements? I’d not argue that they are less probable, as the competition for autonomy and robotics is much stronger than the regular automotive market was in the 2010’s (they were clueless). And also that Musk himself has become a wild card. On balance, I’d say the stock is pretty fairly valued. If the market agreed with you that success in autonomy and robotics was a very low probablility event, the stock would be below $100. And to be fair, some analysts do argue that.
Both autonomy and robotics will be world-changing, disruptive technologies. Autonomy will disrupt the entire transportation market, from trucking to taxis to personally owned vehicles. The use cases for automotive/trucking transport will radically change as economics and convenience undergoe massive re-evaluations. The company that solves those in a financially attractive way will make a lot of money.
Actually, there are Chinese companies doing camera-only for their autonomous efforts: XPeng, AITO (backed by Huawei), Jiyue (backed by Baidu) are three I know about.
Retraining is SOP (Standard Operating Procedure) for all AI-based efforts these days. It’s why Nvidia’s high powered chips are in such demand. Tesla doesn’t necessarily have to gather new data to retrain a modified system, just run the existing dataset through the new system. This is typically a time measured in weeks, not months.
It’s actually more noteworthy that even those companies using LiDAR have been modifying their systems to use fewer LiDAR - and in some cases eliminated the LiDAR from some of their products (eg Mobileye).
It’s not just cost-savings, it’s going to be a revolution in surface street transportation - trucking and personal car ownership are going to look quite different in the future. It would be like saying in 2008 that Uber is just a taxi service using an app and non-professional drivers, and so it won’t change taxi service forever. Go tell that to million dollar medallion owners in NYC.
No, just showing how your arguments on the human senses needed for driving are, in fact, not needed. If you want to now argue that automated driving needs artificial light, go ahead and make that argument.
A significant important fact about the Uber/Lyft services wasn’t just that they were a taxi service with an app. It was that it was asset lite…all the cars were provided by their owners who were gig workers so they could quickly scale up almost anywhere. And they did, And it still took a decade or more to make any money.
And, although Musk has talked about any Tesla joining the Tesla robotaxi fleet, this is not what we are seeing as it plays out at least so far. The cars need to be monitored via remote operators, etc and if a personal car was added to the fleet who would be responsible for charging, cleaning, remote assistance etc. If Tesla does all those things how much profit vs hassle can be left for the owner?
I can see how it might work for a mini fleet operator who owns a few dozen cars, such as rental agencies that loan out their spare cars during low rental demand periods.
Note: I was at a computer vision conference a month ago and there was panel with a Waymo person there. When asked about the costs of their cars the reply dodged this question and instead said the most important cost reduction was with the fleet operations – parking, charging, cleaning, remote operators, assisting stuck cars. These things seemed the most difficult to get to a lower cost.
Bandwidth. I don’t think it’s insurmountable, but running video feeds from 50-100 cars in a city is quite different than running 10,000, and setting up a monitoring center to keep track of them in real time and help the hapless certainly presents some challenges.
Neither taxis nor busses nor Uber nor Lyft have tried to do anything like this. Again, I don’t think it’s insurmountable, but neither do I think it’s a chip shot.
Waymo doesn’t monitor real-time feeds proactively - they wait until a car pings the remote operator to ask for help. Musk has said the same for its robotaxis in Austin.
After the fact, you have the liberty to interpret things differently. Tesla’s dance with bankruptcy was real.
Just to reiterate, I am not questioning whether Tesla can solve these problems, probably yes, but will that probability result in 5x, 10 x payout from current market cap is the question I am not able to answer.
Especially since there is a significant question as to how many others will also do so, whom the leaders will be, and how significant the leadership position.
I think that’s the fairest question. I have no doubt that there will be a business there, but the predictions of “nobody will have to own a car because they’ll all rent them by the trip” strikes me as fanciful. Surely in a few highly congested vertical urban areas that may prove true, but in suburbia? I’m doubtful.
Autonomy success doesn’t mean “nobody will own a car” anymore than TV in the 1950s meant “nobody will go to movies” or that streaming meant “nobody will watch linear TV” or that Cloud Computing meant “no company will have their own data center.”
Yet, those were the right places to place your money.
I know this to not be true since I attended a talk by some company, 2-3 years ago that was installing the cameras in NY city buses in the 2nd phase after the first phase was working well.
And…According to ChatGPT, here is one example:
Gatekeeper has over 3,500 customers, including OEMs for school buses, trains, and transit buses
I think it is reasonable to say lots of transit services are doing this and getting the needed bandwidth isn’t a question. One or two feeds from cameras on a bus are nothing compared to a dozen of the riders all using FaceTime, watching videos, live streaming their influencer videos, etc.
Goofy: I have no doubt that there will be a business there, but the predictions of “nobody will have to own a car because they’ll all rent them by the trip” strikes me as fanciful.
I 100% agree with this. Maybe some robotaxi trips will be replacements for car-owner trips, but many will just be more trips, more miles driven.
I mean, if the easy access and low cost of app-based rides was going to lead to less car ownership wouldn’t we have seen at least some of that due to Uber/Lyft?
I’d like to point out that: since the advent of TV, movie companies (and exhibitors) have been a pretty crappy investment. A few have prospered, many more have disappeared.
With the advent of streaming linear TV is down to about 30% of viewing. Those have not been particularly great businesses either, with the old line networks CBS, NBC, ABC all changing hands, and working in triage mode for decades. (Side note: I remember when Doc Severinson had a 30 piece orchestra for the Tonight Show. Now there are what, 5 guys?)
It’s a bit early to make a call on data centers, but it seems that “cloud computing” has convinced a lot of companies to forgo having their own.
Anyway, I am only responding to the hyperbolic forecasts I have read from the fanbois contingent. When this market does develop, we have no idea how big it will be or how it might play out long term. Will it be individual users and Tesla takes a slice? Or will it be large corporate with Tesla owning the metal and running it like a Hertz? Or a bunch of little entrepreneurs with small fleets? And, for that matter, will it be Tesla only, Tesla and another, or Tesla and a few others?
We just don’t know. Yet the stock seems priced for perfection, and while that has come true in the past (Amazon) it hasn’t always, as the dot-com bust twenty years ago shows oh-so-clearly.
Yeah, that - and the other examples - are my whole point. We don’t need the over-the-top complete replacements of the old to make the new a market-beating investment.
It’s fair to ask if Tesla will have a big enough slice of the autonomy and robotics pies, but not, IMO, to ask if those are two of the giant future pies.
Why not? Sometimes the “giant future pies” turn out to be very small pies, or even turdpies. Let’s look at “building tunnels underground between cities” as one genius proposed and funded. Not a particularly big pie.
I’m remembering “flying cars” as a non-existent pie, and the Metaverse convinced one genius to change the name of his already successful company. The Metaverse, I think, did not take off.
The Segue, Google Glasses, Crystal Pepsi, Fake Meat, BetaMax, LaserDisc, Edsel, and lots of other things were predicted to be huge, potentially industry changing, possibly life changing. And yet, no pie.
Now it may be that Robotaxi will be huge. Or it may merely be big. Or maybe just OK. The industry, thus far, is wildly unprofitable, barely crawling much less walking or running. Yes, there are lots of hopes and dreams, but I submit they are already reflected in the Tesla stock price. If the industry doesn’t develop exactly as Leon hopes, he’ll have to settle for (gasp) being a cool car company. Not a bad end, but not worth what the stock says it is.
You’re stretching here. No-one expected the Edsel to be “huge.” BetaMax was big - VHS just stole its cheese. Much like some expect Tesla to steal Waymo’s.
OTOH, we all remember the Microsoft executive who laughed at the iPhone because it didn’t have a keyboard. And how many people claimed businesses would never trust their data to a third party? And, of course, the whole Electric Vehicle thing, which many didn’t think was going to happen a bit more than a decade ago.
At this point all we can do is bookmark the thread and come back in, say, 5 years.
So, when I look at 2 of the 3, XPeng and Jiyue, I see they use many camera, radar and ultrasonic sensors.
If 2 or more of these sensor types are used to make driving decisions (we can exclude parking, if we’d like), then i would consider that to be multi-sensor and not camera-only (regardless of how this tech is represented in public announcements and branding).
Are you saying that these two companies each have at least one model that only use the camera data to make driving decisions?
Maybe some model or trim is truly camera-only and some are multi-sensor?
My understanding is that Teslas produced today are truly camera-only (excluding map data, which we can say is used for navigation, although I would be interested in how one might define a “driving” vs “navigation” decision - my terminology).