Hacker News Reader

Comment by yalogin | Hacker News Reader

Bear with me here, trying to come at this from the other end.

Why do Uber and Lyft have to develop their own self driving tech? If what they sell is their routing algorithm for drivers, they can simply lease self driving cars from the companies directly and run them on their network. The car companies would be happy to get a cut of the profits.

What advantage would they get by developing their own tech? They have to manufacture the cars themselves or lease the tech to other companies to cover the costs.

Uber and more so Lyft seem late to the game and Uber in particular seem to be fraught with problems and have not made the progress they hoped. So why pursue that? I understand if they started early and are making great progress. On the flip side is the tech no where near completion? Then that would make sense. But Google seems to indicate that they are very very close to launching their own taxi program.

View on HN

Uber and Lyft have very little moat to protect them if someone else develops self-driving cars first.

They haven't succeeded in their initial goal of exterminating the existing cab companies, and that leaves them in a situation where developing self-driving technology could be a matter of life and death for them. The incumbent cab companies are much better positioned to profit from self-driving cars. They've already got all the infrastructure, expertise and staff it takes to manage fleets of vehicles. The bigger ones already have their own apps (e.g., Curb). If self-driving cars hit the general market first, basically all they have to do to run Uber and Lift out of town is buy some and factor the fact that they aren't paying cab drivers anymore into their prices.

By contrast, most of Uber and Lyft's basic business models right now are predicated on the idea of hiring humans, who in turn manage their own cars. That has deep implications: It means they have zero expertise in fleet management. They don't have any of their own garages for storing cars and handling maintenance. They do have apps, but the most valuable parts of their apps are the parts for managing humans. Incentive mechanisms designed to try and ensure that supply of rides and demand for rides remain relatively stable, for example. Their existing business operations are heavily built around recruiting humans - both riders and drivers. None of that is particularly useful in a world with self-driving cars, and the bits that are - basically just the mechanics of enabling people to hail rides and pay for them - are drastically easy to copy.

The first self-driving car deployment could be someone who buys a single car and uses Uber and Lyft as pre-made customer-finding platforms. If self-driving cars become cheaper than humans by enough of a margin for cab companies, then quick movers with capital can work with Uber or Lyft to deploy a self-driving micro-fleet of their own. It would be a great way to park your money, a much cheaper version of buying lease properties or fast food franchises.

Tesla has stated that this is exactly what they would allow customers to do, rent out their self driving cars when not in use to generate revenue for the customer and I assume Tesla would take a cut as well.

But not, I think, cheaper than a cab company setting up their own fleet.

Being a bigger player, a cab company is better-positioned to take advantage of what few economies of scale exist in this space: The power to negotiate a lower per-unit cost, and the ability to operate your own garage so you don't have to pay retail prices for maintenance and repairs.

The cab company would also be the more vertically integrated option, which gives them additional opportunities to economise and pass the savings on to the customer.

This seems mostly right.

That said, worth noting that the fleet management/maintenance expertise will be of unknown/unproven value for managing self-driving cars. No domain experience; just spitballing...

- Unmanned SDC might have significantly different refuel/recharge/cleaning/roadside assistance/accident-response needs. - Unmanned SDC availability/hours/mileage patterns are likely to differ significantly over time when they aren't tethered to bodily functions. - The size and location of fixed fleet-management infrastructure may be suboptimal for different patterns; they might start off as an advantage but become a drag as players with less inertia optimize. - SDC sensor maintenance is going to be a big unknown to these companies. - Their liquidity and vehicle retirement/acquisition processes may not be compatible with rapid fleet replacement without outside help. - Even if they do manage a rapid replacement cycle, the throughput of their fleet management processes may be closely tied to their normal acquisition cycles. - Unless they're going to choose that moment to transition to an app-only-ride-request model, they may have to overcome significant design/software/hardware dev challenges of their own to handle non-app customers (i.e., being hailed on the street, taking verbal directions from people who may be drunk or not know where they're going, stop requests, destination changes...).

Waymo has Avis handling their fleet management. The rental car companies are in a good position to take on Uber/Lyft if they get self driving. They're already renting cars; this is just shorter term rentals.

I can also see big automakers helping their dealerships convert into a network of fleet managers.

I think you are very wrong. Uber and Lyft are better positioned than the taxi companies. They are bigger, have more money, and could even let other people buy the cars, and let them run them. Taxi's have nothing but people, which is something you don't need, and old cars.

> they can simply lease self driving cars from the companies directly

Nobody really knows what company will develop truly practical SDC tech first. Therefore nobody really knows what revenue model that company will have.

If a vehicle manufacturer does it first, then probably their main goal is still to sell equipment, so they would probably be open to selling/leasing it to anyone and everyone.

If a tech company gets there first (or any company without an existing revenue stream), there's always a chance this company will want it all for themselves.

Sure, Uber and Lyft have the name recognition. And they have the active users (accounts created, apps installed, payment methods on file). But SDC is such a killer tech that whoever has it might decide it outweighs everything else and they can build their own app and get the users to come over.

If you're Uber/Lyft and this happens, you could be completely left out. Not only is your service less sexy, you're also paying labor costs that your SDC competitor isn't.

"Nobody really knows what company will develop truly practical SDC tech first."

No, bu the ones with all the patents will be Waymo, Uber, GM.. Lyft?

How does this work? Seems like Waymo would have been able to pick up enough of the obvious ones to kill competition for 20 years. Or just MAD between the big players?

> No, bu the ones with all the patents will be Waymo, Uber, GM.. Lyft?

Considering it looks like were at least 10 years away from having a self driving car industry those patents aren't going to have a long shelf life.

> SDC is such a killer tech that whoever has it might decide it outweighs everything else and they can build their own app and get the users to come over

I'm skeptical. This involves a tremendous capital outlay. Even if you solve that problem, you have to get the production lines running, logistical networks humming, and marketing gears going. Much simpler to buy or partner with one of the companies who have spent years training consumers to press a button and unquestioningly get into a car.

Masayoshi Son is happy to lend you billions of dollars if you have a working SDC implementation.

> billions of dollars

Production isn’t simply a throw-money-at-it problem. And in any case, if you can get billions to build you can get billions to buy.

Mr Musk would like to have a word with you.

> Mr Musk would like to have a word with you

Tesla had much more than massive capital availability. If anything, it spent most of its life at a significant capital deficiency relative to the competition.

For Tesla, there was nothing to buy. The only option was to build.

> What advantage would they get by developing their own tech?

Same with every company in the current climate. Develop every damn thing yourself for fear of anyone having financial power over you.

It really is tiring. Yet, I can't blame them too much. This is where I wish the government would step it. What do I care which company has the best self driving AI? Ffs, define best? It's likely the safest. So we're literally letting them compete for safety - knowing that some will be less safe than others.

There should be a government program here to share knowledge, license startups and ensure equal progression for the benefit of public safety.

I really hate that our safety is being handled by Uber of all companies. Especially since they've already killed someone.

The government currently lets anyone drive until proven unsafe. Self-driving initiatives are really only adding a few "drivers" to the road. Uber acted like a drunk teenager and they are off the road now. This system works, but with occasionally horrific humane toll. It's called capitalism

It's amusing that the California DMV autonomous vehicle testing rules Uber bitched about are very similar to those for a learner's permit:

- Must have a licensed driver in the car.

- Must report accidents.

- No driving for hire.

- No heavy trucks.

It's not the only capitalistic way to do it. You could have private companies developing the technology but a government testing program to demonstrate some level of safety.

That is what they do with impact crash testing. Private companies design cars, and then the government crashes some to see how they do in certain scenarios.

But, the US government isn't taking this route right now, probably because they believe that SDCs have the promise of reducing traffic fatalities over the long term, so they don't want regulation to delay the process. But that's a policy decision based on calculated risk, not a necessity of a capitalist system.

> That is what they do with impact crash testing

Eventually

https://en.wikipedia.org/wiki/Unsafe_at_Any_Speed

I stand by my assertion that this is just how capitalism works

> The government currently lets anyone drive until proven unsafe

I mean, sort of - you can't drive unless you pass a test of safety. I'm not sure exactly how that fits your model of "until proven unsafe".

Driving without a drivers safety course having passed, ie a drivers license is a serious offense.

> I'm not sure exactly how that fits your model

It only takes a written test to get a permit. You don't have to demonstrate safety, only knowledge of the rules of the road. You also need an adult over 21 years old to sit in the passenger seat. It is the lowest of bars

Sure but that's not a permanent thing, and frankly that is comparable to the current testing model of cars.

Furthermore, humans are not crafted beings. We can't cooperatively construct the knowledge of a young driver. Why is the terrible human bar our basis for success? Especially since this bar is only so low so that we can allow companies to compete? To have financial freedom to kill people?

My argument is that public safety does not need to be driven by financial success of companies. We've got the chance here to share knowledge for the common good. Yet, we're holding onto it so that companies can have the chance to make money.

Furthermore, why is "the lowest of bars" (in your words) our metric for success here? Usually the lowest of bars is a bad thing.

Humans are terrible abysmal drivers. I want them off of the road asap. Yet, I don't think hoarding public safety related data for financial success is the best effort towards achieving our goals.

This is a silly game. If self driving tech can even work (questionable in the next few decades), it’ll immediately become a commodity. No such tech could be protected for long. There’s not going to be huge profits to be made once multiple players have self driving fleets on the road. And since unattended cars will inevitably become trashed, individuals will just buy their own self driving cars, and traffic will get that much worse.

I can't imagine there won't be huge patent fights. Whatever is needed to make self driving work is clearly non-obvious or it would be done by now.

It may be the case of just throw enough ideas together and that works, in which case it may be effectively a commodity; everybody working on it has enough patents that cross-licensing works. Or it might be there's one key idea that makes everything work, and whoever files first controls it for whatever a patent lasts these days. Kind of depends which company owns that patent how it goes.

From what I can tell, all of the other self-driving companies plan to launch a taxi service first. Even Tesla has plans for that, as evidenced by their "no lyft/uber self-driving" clause in their TOS for any Tesla you buy.

I didnt know this. I see plenty of teslas on uber.

Only for the hypothetical self driving mode.

And yet again head over heels. Most Teslas made today won't even be around to drive when SDC becomes a reality.

I doubt Lyft wants to sit around and let Uber develop self-driving tech, since I doubt Uber would lease that tech to Lyft on very favorable terms.

I doubt Uber wants to sit around and let Lyft develop self-driving tech, since I doubt Lyft would lease that tech to Uber on very favorable terms.

I have just read at https://www.lemberglaw.com/self-driving-autonomous-car-accid... a little bit about this topic. I think it's always interesting to talk about this future technology. However, for now, I wouldn't risk my family to ride inside one of these self-driving cars. They haven't 100% perfect/safe. Maybe 10 more years.

I imagine Uber+Lyft don't consider their driver routing algorithm a significant barrier to entry to the self-driving car market.

They likely expect the first company to come up with true self driving cars to start their own competitive service to Uber+Lyft. To them, it's an existential threat to not invest in their own self driving technology.

I think every self-driving car project is under the impression that the first stable and safe self driving car network is going to have no incentive to lease the technology out.

Yeah, making an Uber-or-Lyft-like service is easy. Lots of people have done it. Uber and Lyft price-competed those services into irrelevance, but if another service has self-driving cars and Uber-or-Lyft don't, then they won't be able to price-compete effectively.

I think it's likely that if Uber-or-Lyft don't get a sustainable technological advantage with driverless cars, they will actually stick around and make a deal with the company that provides the driverless cars, but they will have a weak negotiating position and will have to accept a small percentage of the profit which will not justify their current valuations (especially Uber, which has a much higher valuation than Lyft).

You are right. There is no good reason for them to develop and build autonomous cars, just like there is no good reason for them to develop and build conventional cars. When autonomous cars take off, Lyft and Uber can still use their current model. The only difference would be that the car owner doesn't have to drive.

I'm gonna have to side with you here, I don't see the point of what they're doing. Except for distinction and marketing. Distinction, as in, if you have total control of how your car, it is easier to create functionality that competitors don't necessarily have. Marketing, as in, marketing.

What incentive does a car company have to lend them the technology vs running self driving taxis themselves?

By that logic why not run their own taxi and car rental companies today? Vertical integration can make sense but it also often doesn’t.

Reply on news.ycombinator.com