-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
- Status
adiggs
Well-Known Member
At least one big question here @tlo is what the uptake will be on this competing autonomous fleet. As you say, TSLA can achieve scale on availability comparable or better to UBER's for not much money (or at least what sounds like a readily achieveable quantity of money considering the market value of UBER, and the possibility of competing or replacing).
This isn't really a TSLA specific question - it's a human adoption question, with like any of these questions, a range of answers for people. Some people won't consider getting into a car with no human pilot for years or decades to come. Some people will stand in line to be the first person in a car with no driver. Will there be enough riders for TSLA's ride sharing initiative to show revenue and value? Will they show up fast enough to overcome UBER's network, and establish a new network?
Heh - does UBER look at the same math, and decide to buy TSLA?
If the uptake on the autonomous fleet is strong, then everybody that doesn't have autonomous cars is sunk. And TSLA is overnight the most valuable property in the world depending on the gap between when TSLA is producing at scale (10's or 100's of thousands/year), and when the next company is at the same scale.
Can a gas engine based car achieve comparable autonomous performance? If it can, then there will be the possibility of competition soon after. If not, then everybody else is a full gigafactory worth of battery production behind (oops
).
I challenge the unstated assumption though, that an initial solution to autonomous driving, makes this a solved problem. I expect we'll need at least 2 additional major iterations from first major release of autonomous driving for it to be reasonably complete and usable. Consider that with all of the tweaks TSLA has implemented around auto pilot so far over the last year, I would classify everything in production up to today as still being version 1.0 of adaptive cruise control and lane keeping. Still not v1.0 of autonomous. From what I read of FW 8.0 with highway to highway transfers via on-ramps, that sounds like v2 of TACC plus lane keeping.
In reality, I expect we'll see a full decade of obvious and noticeable improvements over the initial autonomous functionality. This might provide a window for UBER and others to catch up. And it might prove to be such a huge mound of a learning challenge, that every passing year TSLA is climbing that mountain, the further behind everybody else falls. (As an investor in this, I claim that we want it to be hard - if it's easy, then everybody will do it).
This isn't really a TSLA specific question - it's a human adoption question, with like any of these questions, a range of answers for people. Some people won't consider getting into a car with no human pilot for years or decades to come. Some people will stand in line to be the first person in a car with no driver. Will there be enough riders for TSLA's ride sharing initiative to show revenue and value? Will they show up fast enough to overcome UBER's network, and establish a new network?
Heh - does UBER look at the same math, and decide to buy TSLA?
If the uptake on the autonomous fleet is strong, then everybody that doesn't have autonomous cars is sunk. And TSLA is overnight the most valuable property in the world depending on the gap between when TSLA is producing at scale (10's or 100's of thousands/year), and when the next company is at the same scale.
Can a gas engine based car achieve comparable autonomous performance? If it can, then there will be the possibility of competition soon after. If not, then everybody else is a full gigafactory worth of battery production behind (oops
).I challenge the unstated assumption though, that an initial solution to autonomous driving, makes this a solved problem. I expect we'll need at least 2 additional major iterations from first major release of autonomous driving for it to be reasonably complete and usable. Consider that with all of the tweaks TSLA has implemented around auto pilot so far over the last year, I would classify everything in production up to today as still being version 1.0 of adaptive cruise control and lane keeping. Still not v1.0 of autonomous. From what I read of FW 8.0 with highway to highway transfers via on-ramps, that sounds like v2 of TACC plus lane keeping.
In reality, I expect we'll see a full decade of obvious and noticeable improvements over the initial autonomous functionality. This might provide a window for UBER and others to catch up. And it might prove to be such a huge mound of a learning challenge, that every passing year TSLA is climbing that mountain, the further behind everybody else falls. (As an investor in this, I claim that we want it to be hard - if it's easy, then everybody will do it).
Google Takes on Uber With New Ride-Share Service - The Wall Street Journal
Google Takes on Uber With New Ride-Share Service
Google to bring Waze ridesharing San Francisco-wide, may set stage for driverless service
Google Takes on Uber With New Ride-Share Service
Google to bring Waze ridesharing San Francisco-wide, may set stage for driverless service
Last edited:
MitchJi
Trying to learn kindness, patience & forgiveness
At least one big question here @tlo is what the uptake will be on this competing autonomous fleet. As you say, TSLA can achieve scale on availability comparable or better to UBER's for not much money (or at least what sounds like a readily achieveable quantity of money considering the market value of UBER, and the possibility of competing or replacing).
This isn't really a TSLA specific question - it's a human adoption question, with like any of these questions, a range of answers for people. Some people won't consider getting into a car with no human pilot for years or decades to come. Some people will stand in line to be the first person in a car with no driver. Will there be enough riders for TSLA's ride sharing initiative to show revenue and value? Will they show up fast enough to overcome UBER's network, and establish a new network?
Heh - does UBER look at the same math, and decide to buy TSLA?
If the uptake on the autonomous fleet is strong, then everybody that doesn't have autonomous cars is sunk. And TSLA is overnight the most valuable property in the world depending on the gap between when TSLA is producing at scale (10's or 100's of thousands/year), and when the next company is at the same scale.
Can a gas engine based car achieve comparable autonomous performance? If it can, then there will be the possibility of competition soon after. If not, then everybody else is a full gigafactory worth of battery production behind (oops
I challenge the unstated assumption though, that an initial solution to autonomous driving, makes this a solved problem. I expect we'll need at least 2 additional major iterations from first major release of autonomous driving for it to be reasonably complete and usable. Consider that with all of the tweaks TSLA has implemented around auto pilot so far over the last year, I would classify everything in production up to today as still being version 1.0 of adaptive cruise control and lane keeping. Still not v1.0 of autonomous. From what I read of FW 8.0 with highway to highway transfers via on-ramps, that sounds like v2 of TACC plus lane keeping.
In reality, I expect we'll see a full decade of obvious and noticeable improvements over the initial autonomous functionality. This might provide a window for UBER and others to catch up. And it might prove to be such a huge mound of a learning challenge, that every passing year TSLA is climbing that mountain, the further behind everybody else falls. (As an investor in this, I claim that we want it to be hard - if it's easy, then everybody will do it).
The time frime for full autonomy in inner cities is going to be awhile. Look at the subway in NYC, even that is not fully autonomous(when it should be).
adiggs
Well-Known Member
I understand that, and if we interpret this to mean signs of life in the wild in 2 years, then we're saying the same thing.
It doesn't mean I'm right, but bringing autonomous driving out of the various pilots, test labs, and other incubating environments and into the real world for the first time is one thing. Expanding it from there to the wide variety of corner cases that exist only within a city environment in the United States is another.
Expanding from there to the wide variety of different corner cases that exist in progressively more rural and less paved environments is another thing.
Expanding from these to other countries that are equally good or better about organized traffic patterns is another.
Expanding to yet other countries where traffic less clearly obeys laws and follows patterns that are consistently and repeatably recognized by the autonomous driving software is another.
In many ways, the initial release of autonomous driving will also mark the most difficult environment in which autonomous driving software will need to function. With enough elapsed time and familiarity with the technology, enough optimizations to how we mark our roads and signs, add sensors and electronic information and communication to our roadways, and otherwise adjust our infrastructure from a human based to a machine based infrastructure, it'll become harder and harder for a human to function as a driver.
I agree with Elon. I believe he would agree (and if not, then we disagree on this point) - first release of usable software marks the beginning of a whole mound of learning, regulatory change, infrastructure change, and cultural change; not the end. We aren't 2 years from having this autonomous driving thing fully implemented. We're 2 years from the start of the implementation and change.
My prediction / guess, is that the implementation and change is going to drag out for a couple of decades.
As an investor, my prediction is that the direction the breeze is blowing will become apparently within a few years of initial introduction, with the remainder of the couple of decades devoted to the massive task of making the changes.
@tlo It's a good exercise... here are some of my thoughts.
First, I think you're comparing the ride-sharing/Uber's market to the taxi market and I think that's incorrect. Uber is in a different market. In fact, I'd say Uber's addressable market is at least 10-50x as big as the current taxi market. The reason being is that Uber is causing people who normally wouldn't use taxis to use their service. One example, I'll drop off my car at a shop and take Uber home. Without Uber I wouldn't do that, I might wait around at the shop or do something else. Or another example, I might have my wife drop me off somewhere and tell her I'll take Uber and meet her in a couple hours. Taxis aren't prevalent where I live, so I wouldn't do this kind of behavior unless there was Uber (or a ride-sharing service). In cities, it's even more extreme. People are starting to ditch their cars and exclusively take Uber. They normally wouldn't even think or entertain that possibility in a world of just taxis. But Uber/Lyft are so reliable and convenient that it allows people to use ride-sharing service much more frequently than they would use a taxi. This is what is fueling the hyper-growth of Uber.
So if there are 250k taxis in the U.S. currently, then we'd need to calculate how big Uber's market would be by the time fully autonomous vehicles are approved by regulators. My guesstimate is 2022 (as I've shared in a previous post).
In that case, by 2022, how big will Uber be in the U.S.? I don't know. But I think it's safe to say they'll be a lot bigger than they are now. Especially since they're still growing at a 2-3x clip (although I'm not sure exactly what their growth rate in the U.S. is). I wouldn't be surprised if they're at least 10x as big as they are now in 6 years... They could be up to 20x as big. So, let's say current day Uber is equivalent to the taxi volume in the U.S. (even though I think it's already eclipsed it) at 250k cars. In 6 years, Uber could be at 2.5M to 5M cars (that's 10x-20x).
At 2.5M cars, it's going to take quite a bit to displace Uber.
Sure autonomous cars are going to be much more efficient than cars w/drivers. So, it's possible that 1M or less autonomous cars could replace 2.5M+ cars with drivers.
But even 1M autonomous cars is a lot, and that's just the U.S.
If Tesla were to own the 1M autonomous cars, they'd need to invest at least $30+ billion. That's a lot of money. I don't know where Tesla would get that (unless they have some crazy market cap by that time and can raise that money).
Another option would be for Tesla to have individuals purchase the autonomous cars and have the autonomous cars be placed in Tesla's sharing network.
Here's how I think Tesla has a chance to make this work.
Tesla could PRIORITIZE orders from people who agree to the following:
1. They must live in one of the cities that Tesla is targeting w/ride-sharing services.
2. They must agree to share their car to ONLY Tesla's ride-sharing network when they're not using it or planning to use it.
If demand is high for Tesla's autonomous cars, perhaps many people will agree to the two conditions listed above. Tesla also could give a discount for those who agree to the above conditions (or charge higher to those who don't agree to the above conditions).
If Tesla's production is high enough, and they get enough people who will agree to the above two conditions, then Tesla can deploy autonomous vehicles in high numbers in targeted cities. These autonomous vehicles would be owned by individuals and those individuals will have agreed to share ONLY to the Tesla network.
If Tesla can successfully pull this off, and if by 2022 they are making let's say 2M cars a year, then they have the chance to disrupt Uber. Perhaps Tesla can get at least 1M autonomous cars out in 2022 this way. And if Uber is 1 year late with autonomous cars (ie., by partnership with other manufacturers), then this gives Tesla a way in to compete with Uber. If Uber is head-to-head with Tesla and releases autonomous cars at the same time as Tesla, then Uber can release million(s) of autonomous cars at the same time or even faster than Tesla (since Uber would be partnering w/potentially many auto makers). In that case it'll be much more difficult for Tesla to beat Uber, since Uber has their existing network and they're only making it better.
Tesla would still have a chance though because perhaps their cars would be better (ie., safer, more comfortable, etc). And that could be a selling point for Tesla's network.
Another point to make is that by 2022, it could be a messy and ultra-competitive market. Especially if Google and Apple get into the game as well. It's hard to imagine what kind of crazy fight it would be if Google, Apple, Uber and Tesla all got into the same ring. And in other countries there are formidable companies as well.
Everyone (ie., Google, Apple, Tesla, Uber, etc) knows transport is a huge multi-trillion dollar market, and it's going to be disrupted. And they want a big piece of the pie.
Who's going to make Uber's EVs and how will they ramp as fast as Tesla (by 2019 lets say) without one or more Gigafactories?
Uber isn't silly enough to go self-driving with ICE cars are they? That would be suicide from a cost/maintenance point of view. Not to mention the fact that gas/oil is due for an imminent disruption event once the world fleet starts to electrify.
If every Model 3 is capable of eventually becoming a self-driving taxi, and the service takes form in the 2019/2020 timeframe, then Tesla should be in a great position
Google's the biggest puzzle of all. I imagined they would turn their self-driving tech into Android Auto, such that it would be the self-driving operating system for any/every car. But, if they shopped it around it appears no one was biting. And now, their talent is bailing for other companies and startups. Really surprising to see them flat-footed when they appeared to be years ahead in this area.
Uber isn't silly enough to go self-driving with ICE cars are they? That would be suicide from a cost/maintenance point of view. Not to mention the fact that gas/oil is due for an imminent disruption event once the world fleet starts to electrify.
If every Model 3 is capable of eventually becoming a self-driving taxi, and the service takes form in the 2019/2020 timeframe, then Tesla should be in a great position
Google's the biggest puzzle of all. I imagined they would turn their self-driving tech into Android Auto, such that it would be the self-driving operating system for any/every car. But, if they shopped it around it appears no one was biting. And now, their talent is bailing for other companies and startups. Really surprising to see them flat-footed when they appeared to be years ahead in this area.
A good read about Uber vs Google vs Tesla,
Google, Uber, and the Evolution of Transportation-as-a-Service
The author seems to count out Tesla... which I don't agree with.
Google, Uber, and the Evolution of Transportation-as-a-Service
The author seems to count out Tesla... which I don't agree with.
This is probably the best article on Amazon that I've read in a while. It details Amazon's ridiculous growth the past several years and their ambitious inroads into all things transport.
Will Amazon Kill FedEx?
I bring this article up because I think it might be only a matter of time before Amazon starts exploring transporting people as a service as well. A few months ago there were reports of Fiat talking with Uber and Amazon about partnering for autonomous vehicles.
Will Amazon Kill FedEx?
I bring this article up because I think it might be only a matter of time before Amazon starts exploring transporting people as a service as well. A few months ago there were reports of Fiat talking with Uber and Amazon about partnering for autonomous vehicles.
Interesting, thanks.
Here's a question - If others besides Tesla do actually create software for autonomous cars and they are working at the same time, what happens when Tesla's software works better? If people get in Uber's or Google's cars and it doesn't function as well, I would think customers would jump ship and start using Tesla. Or they could sale their old car and buy a Model 3 and then rent it out.
Example - The article talks about Uber's mapping, but Tesla will probably have over a billion miles of autonomy by the time this all plays out as competition.
Last edited:
Interesting article about people in LA starting to ditch their cars in favor of Uber/Lyft,
People in Los Angeles Are Getting Rid Of Their Cars
"Spiegelman had been studying the economics of riding Uber and Lyft versus a taxi or driving a personal vehicle when he decided to run the math for his own car. He made a spreadsheet outlining the cost of leasing his Volkswagen: $458 monthly for the lease itself, $158 for insurance, $70 for gas, and at least $72 for parking, for a total cost of about $758. Based on those calculations, he said he has saved more than $1,100 in the last three months, spending an average of $3.42 for each UberPool or Lyft Line ride to work in August."
This is why you can't compare the Uber's addressable market to the taxi industry. The taxi industry (due to high costs and high inconvenience) doesn't tempt people to ditch the cars in masse.
As more and more people ditch their cars, this increases Uber's addressable market considerably because now these people almost completely rely on Uber/Lyft for their transport needs (of course it depends where they live).
People in Los Angeles Are Getting Rid Of Their Cars
"Spiegelman had been studying the economics of riding Uber and Lyft versus a taxi or driving a personal vehicle when he decided to run the math for his own car. He made a spreadsheet outlining the cost of leasing his Volkswagen: $458 monthly for the lease itself, $158 for insurance, $70 for gas, and at least $72 for parking, for a total cost of about $758. Based on those calculations, he said he has saved more than $1,100 in the last three months, spending an average of $3.42 for each UberPool or Lyft Line ride to work in August."
This is why you can't compare the Uber's addressable market to the taxi industry. The taxi industry (due to high costs and high inconvenience) doesn't tempt people to ditch the cars in masse.
As more and more people ditch their cars, this increases Uber's addressable market considerably because now these people almost completely rely on Uber/Lyft for their transport needs (of course it depends where they live).
If Tesla's autonomous cars work better it'll be a big competitive advantage. However, there's also a lot of competition as some of the smartest tech companies in the world are working on autonomous driving... like Google, Baidu, probably Apple, Uber, and others as well like Mobileye, Cruise, Comma.ai, and a whole host of others. But Tesla does have the advantage currently as they are deploying cars on the road with Autopilot tech and are aggressively rolling out new capabilities. The way I see it, Tesla is in the lead and if they execute well they can keep their lead. But the lead might grow smaller if other tech companies get serious (ie., Google, etc).
palmer_md
Member
Fully autonomous driving will need an integrated system where cars communicate with each other. Cars are connected to the grid and can be linked to traffic lights and traffic routing. It will require coordination with cities, transportation networks and the cars themselves to be fully autonomous. Knowing where road construction is, how fast traffic is moving ahead, how many cars are in front would allow for full autonomy. There will always be a need for a driver in bad weather (snow storms, etc.)
Musk mentioned in one of the previous conference calls that eventually he envisions a system where cars will mostly travel in tunnels(which would be perfect for autonomous driving - a controlled environment, no sunlight to throw off sensors, no snow, no intrusions in the road).
Before we get to full autonomy, we need a connected grid where cars can see miles ahead, route themselves based on traffic and other car speeds and detect problems like construction or accidents on the road. This wouldn't really work in rural areas but in countries with really developed transportation networks, an integrated system with cameras on every single road, linked traffic signals, and car road speeds would allow for this.
A mass network like the traffic infrastructure in Seoul would speed up autonomy because you could have cars communicate with the entire network(to detect accidents in advance - like the crash in China)
Attachments
Last edited:
MitchJi
Trying to learn kindness, patience & forgiveness
At one time it was believed that full autonomy could only be done with some kind of external infrastructure, but over the last few years it's been proven that it isn't necessary.
One of the biggest problems facing full autonomy is handling legacy cars. How does a fully autonomous car deal with a car that was built before anyone was even thinking about the kinds of computing cars have today? The average age of a car in the US is 11 years. That means even if every new car starting today was fully autonomous half the cars on the road would still be pre-autonomous a decade from now.
It's not a super difficult problem to have fully autonomous cars in a world where every vehicle is autonomous and talking to one another as well as tied into a network that is telling every car on the road what is going on down to fine details (more than humans can get from traffic reports today). The problem gets very messy when you throw in even a handful of non-autonomous vehicles. And having a data network to plug into is nice, but what does the car do if it drives into an area where the network hasn't been extended yet? And what does the car do if the support network goes down either through accident or malice?
Early in my career I painted myself into a bit of a corner with a hardware design project when I gave a lot of thought to how it would run when up and running, but forgot to put much thought into how the system would boot and get there. I had to go back to the drawing board to rethink some things that worked fine once everything was booted, but it turned out to be difficult to get them into that state to begin with.
Edge conditions are always a pain, and transitioning from one tech to another can be a major edge condition. In the computer world, Apple likes to attack Windows for being overly complex, but Microsoft put a lot of effort into making Windows backwardly compatible with older software and even older hardware. Apple's approach is to just declare some older stuff obsolete and you can't use it any more. It makes MacOS simpler than Windows, but it also means Macs don't play as well with legacy products.
The road system will have to take more of the Windows approach to legacy cars than the Apple approach because most of those people with cars 11+ years old are driving the old car because they can't afford anything newer. You don't want to force a lot of working poor out of their jobs because they can't afford to get to work anymore.
Autonomous cars will allow some people driving older cars to get rid of their car and take some kind of taxi-like service, but it won't enable all of them. There are many reasons people are driving older cars and no matter what we do, we will have some older cars on the road for more than a decade after autonomous cars are introduced.
Agreed. Further, terms like "always" and it's counterpart "never" are trigger words to scientists and engineers. Always and Never represent longer timeframes than most people comprehend. Infinite time is implied, when 5-10 years is what's meant, and even then might not be accurate.
Simple computer controls have exceeded human capability for decades in wet/snowy conditions for both traction and braking. Further, radar can see through snow, unlike humans. Full autonomy is not only inevitable, but will ultimately be better than humans in all conditions.
An aside on comma.ai -- George Hotz has no plans to enter the Level 4 race. He's solely focused on Level 3 as an aftermarket product, and in fact, refuses to answer when he thinks Level 4 will be possible, generally shrugging his shoulders as if he doesn't care. So, he shouldn't be considered a competitor for any driverless technologies.
Lastly, back to Uber. Julian Cox had an interesting article a couple of months ago discussing the merits of driverless ICE cars Tesla Model 3 & The Economics Of Autonomy — Why An Autonomous ICE Vehicle Is Relatively Pointless where he concluded that the cost per mile would make them somewhat meaningless. It's hard to predict how all this will shake out, and whether L4 autonomy will be both ready and approved before Uber has access to 500,000 EVs, but if we reach approved L4 before that point, it would seem to grant Tesla an advantageous position and leave Uber with a decision of whether or not to deploy L4 ICE vehicles.
- Status
Similar threads
- Locked
