Waymo teams up with Jaguar to intro a new, premium self-driving car
techcrunch.com
techcrunch.com
"As standard the car comes with a 90kWH battery pack which produces a total system output of 400PS and 696 NM.
It is powered by two electric motors which produce 200PS and 348Nm each and are mounted to both the front and rear axles, enabling four-wheel drive.
Due to the instant torque of an electric powertrain the car will be able to sprint from 0-60mph in just 4.5 seconds and on to a top speed of 124 mph.
The car will be able to travel up to 298 miles of range on a single charge."
[1] https://www.express.co.uk/life-style/cars/937715/Jaguar-Waym...
(edit: link)
I chuckled a bit trying to imagine a situation where the SDC logic would conclude that it was necessary to floor the accelerator up to 124MPH.
Well, if every Tom, Dick and Rajiv thinks it's being driven by a computer and they can cut it off ...
I'd guess these cars are being designed to have the self driving system as an addon that can be removed/redesigned in the future as the tech evolves.
The iPace weights just 300KG more than the F-Pace, so it’s not down to extra weight alone.
Will they avoid potholes? In New England, we have MANY bad potholes, I wonder what they will do about that. Will they avoid manhole covers, objects on the road? I personally can't wait to see when/how it all plays out.
This sentence feels marketing more than informing. Whats the initial release volume? Where etc. Up to 20k cars doesn't mean much and they think it will average 50 trips per car per day - seems hopeful.
A little disappointed in this announcement.
Although, if 10,000 were definitely enough, they might not have made an agreement for 20,000, so maybe it does sort of tell you something.
I mean - I think all this talk of going live in 2018 should be stopped by officials until they create a TEST for SDVs. Also, it wouldn't be a bad thing for SDVs to run an internal offline unit test given a fake lidar scenario and make sure it doesn't kill people (every time before the car leaves the parking spot).
He/She is not saying that.
They're saying that the bad PR around Ubers accident means they should have moved this event to a later date.
???
It's one of the perpetual problems of regulation. As soon as you define a test or requirement companies will start gaming the system to meet them instead of addressing the purpose behind the regulation.
The test could have basic things like a car stopped in the middle of the road, a fake person that is fallen on the ground, a tree on the road, a big stone on the road, a big hole in the road and other similar tests.
You can change the obstacle types and sizes so tests differ. The company will not be allowed to use a different software fort this tests so at least you can test that the sensors are good enough and that the software can handle basic things.
This tests would be like the medical tests a driver must pass before starting the driving school, we can trust Uber like companies to perform this tests inside before putting the cars on the road for public road testing.
If the tests are too basic I am sure clever people can find even better tests to stress the sensors and software before allowing the dangerous tests on public roads.
I mean heck, this latest incident wouldn't have passed the original Darpa challenge that got us here.
On the other hand, any self driving company is running an extremely complicated and extensive test suite on their code. Developing these tests is a core engineering challenge to making a sdv and any test developed by a regulator will be a joke by comparison (unless we give the regulator billions of dollars to essentially replicate the work.)
Maybe you could start with a generic test, for instance: x miles in random location with random conditions.
After that you could add tests when a problem appear. For instance, you could add the conditions where that woman was killed the other day. When another accident happens you would add a test appropriated to that conditions.
Regression testing for driving.
The point is to have a documented standard to compare automated systems by.
Move fast and break things works fine when the at risk subject is someone's vacation photos from 2015 rather than a kid in a crosswalk.
By the way, that's 20k from Jaguar. Don't forget the thousands of Chrysler minivans that they ordered.
You can mitigate it a bit by simulating many rides in virtual reality--which Waymo is absolutely doing already--but add enough cities and that would get really expensive really quickly, even by Google standards. And you probably couldn't run that world suite of tests for every code submission, or even on a nightly candidate.
You have a neural network which inspects the code change, and predicts which tests/scenarios might be affected (a change in the vision system is unlikely to affect the handbrake tests for example). You run only the most likely to fail tests. A test which hasn't been run for a long time is more likely to fail, since more code changes were done since it was last tested, so all tests end up being run on a semi-random schedule.
When any test fails, you then run back through time, bisecting every revision till you find the code commit which caused the failure.
At that point, depending on your engineering workflow, you either create a ticket to an engineer to fix this commit to pass the test before the next release, or you auto-revert that commit.
An autonomous car is inherently a system, and you need a deep suite of integration tests, regularly run and covering all possible cases, to believe you're even approaching vehicle safety. The wider you space out when any given integration test is run, the harder it becomes to fix. And if a given test only runs once per week, sooner or later (sooner with a sufficiently large test suite) you're going to hit a place where auto-reverting the commit breaks additional tests.
And when that happens, unless you put a global lock on new code entering the system, it might be impossible to fix the regression. And putting a global lock on new code submission will make new feature development extremely slow and expensive.
I'd be curious to see if a lot of people did a similar thing (giving up their cars). Price is the kicker here though. When I own a car, I can pretty much drive as frequently or infrequently as I want. Where as with Taxi's/etc, the cost would quickly rack up. So these AIs are going to have to be cheap by comparison.
Man, the future of this tech is so exciting for me.
First of all, what is a lot? 1% of the US would be around 3 million people, which is a lot of people and yet 1% is not very significant. Of course we know some people don't own a car, you can make your numbers look bigger or smaller depending on if you count them. (there are books on how to lie with statistics if you want to continue this line)
There are people who are already close to giving up their car who will.
I believe that most people will not give up their own car. Rush hour is a thing for a reason: most cars are driving during those hours, so your costs for a self driving car ride rental needs to account for that. Once your costs are not much different the utility of your own car is a price worth paying. I know that I can go shopping over lunch and leave my new toys in the trunk. I keep a spare diaper in my car just in case my kid needs one. When I go on vacation I often start packing the car a few days before.
If you are frugal rides in a self driving car is probably more expensive than rides in your own car. Your payment has to cover oil changes (I change my own oil - I've concluded that it actually saves me time over driving to a mechanic). Your payment covers inspection/cleaning after every ride - nobody cleans their personal car that often. People who ride in a self driving car will demand newer cars even though a 12 year old car (if well maintained) works perfectly fine for the same task.
Same thing with the car age. They aren't going to cycle new cars in every few years. At best, they are going to replace the interiors in that time frame.
It also doesn't account for the size of the car. People buy 4 seaters for the 10% (or whartever) of the time that they have passengers. Commuters will be perfectly fine with a 1 seater if they can order a bigger car when needed.
They're going to recycle the cars faster than today, because they'll be used for closer to 100% of the time, rather than 10% today.
The auto rental industry does. The average holding time in the rental car industry is 13 months. Enterprise Holdings (Enterprise, Alamo, and National) owns over a million cars. Lots of companies talk about "transportation as a service". Those companies do it.
Among other things, they own enough parking spaces to park all those cars.
Also, if the taxi fleet knows where everyone is going, they can easily organize a car-pool. If you're frugal then you'll be okay with pooling, pay more if you want the exclusivity (or, if you have the frugal plan, you can pay 10 bucks more this morning if you want to be left alone...).
The Jaguar vs Tata would make a fun algorithm. Say you're the frugal one, you order a car, and a Jaguar is heading your way. Suddenly your neighbor (on the premium plan) orders a car, and someone else a few minutes away cancelled their Tata. Oops, sorry, he gets the Jaguar.
Edit: Changed "Mercedes" to "Jaguar"
I don't see any reason to think a similar price-sensitivity wouldn't apply to other forms of seat-rented transportation.
Seriously I do wonder, though, if the design of townships like yours will be directly affected by the rise of AV's. It would seem to make sense for Waymo and others to effectively lease depots (holding pens?) for when you and your fellow travelers need to hail a car. Are you/we effectively trading your individual garage space for a collective AV parking lot on the outskirts of town? And whether that development would compare favorably to investing in more intercity trains?
I don't understand why everyone wants to live in these horrible suburbs with grass you have to mow and a pool that every single person needs to build. I seriously have never once understood suburbs. I get living on 5+ acres, you got stuff you can build and do (tree houses, chicken coups, etc) but what is the point of putting yourself somewhere with nowhere you can walk and nowhere you can build?
Also, can we talk about grass for a second? Why? Just why do we do it at all? Trees, shrubs, and flowers are so much nicer and easier, why do we need to make bright green rectangles with most of our lawns?
The appeal of the suburbs is pretty obvious. There's enough room for you to build a bigger house, some out structures like pools and sheds, and some grass for your kids/dog to play in. The downside of that is it isn't walkable. That downside might not make it worth it for you, but it's worth it for others.
I think for a lot of people the appeal of the suburbs is mostly "I could get a mortgage and I can commute to my job". Living in walkable areas in cities is really expensive (maybe because lots of people want to do it but they've been illegal to build for most of the last century?)
I'm looking at homes and as housing prices have gone up 60% over the last 5 years where I am it's a question of "uggg do I just keep paying rent in the city centre forever or do I buy a place where I can afford one, which is mostly shittacular suburbs". I think a lot of other people are there by necessity, not choice, as well.
The GP makes a good point actually because in that background to this decision there's always "screw this, I'll just move way out in the middle of nowhere and get a bunch of cheap land and a giant house, and work remote" but having a spouse who can't work remote affects that. The suburbs are the worst of both worlds.
As a very much secondary benefit, I get some exercise and a minor outdoor hobby.
Your local fire marshal might have something to say about that.
Money. Lack of.
In lots of US metro areas, the suburbs are the only route for working and middle classes to own property and hopefully pay it off before or by retirement, and only have to worry about property taxes (capped for over-65 homeowners in some states) and maintenance thereafter.
Where in-town living is not more expensive than suburbs, city politics driving assessments plus millage rates and aging building maintenance of generally crappy US construction practices make for a much more volatile financial picture long-term for retired urban dwellers.
In a suburb, on their own lot, it feels like they have more control because they don't have to negotiate as part of a group or even participate within a group for basic maintenance, and property taxes are generally not as high per unit area. Even crappy construction practices in SFH's can be remedied for all but the most egregious cases. When a contractor screws up a boiler installation in a high-rise condo, it's a massive capital investment and logistical headache for all the residents when that boiler calls it quits early, especially if replacement requires some demo to get at and extract the boiler, DIY is not an option, and pipefitters are not cheap. A home hot water heater replacement isn't a financial picnic for working and middle class homeowners, but it is still within the DIY realm for those willing to put in the time to save precious capital.
The operating environment looking forward for the next few decades in the US is most of the population is scarce on capital but have time (for example, retired working and middle class), a substantial portion are scarce on capital and time (for example, households with children), and any solutions (like moving more people into urban areas) that involve capital they don't have are non-starters. Yes, we should be encouraging people to move into urban areas in the US, because the per capita energy expenditures are much lower, environmental impact per capita are lower, people are given greater opportunities to form and maintain communities, etc., but the way incentives are currently structured in the US prevent that from starting.
Just by participating on HN statistically makes it likely you and I are in a rarefied environment where we don't have our noses rubbed into that reality I described above every single day, but we're a tiny, tiny fraction of the US, not to mention humanity, and moving the needle to make a difference in the world requires that we figure out solutions that sit on the other side of that fraction.
if you've ever actually sat down and priced out how much each usage of your car costs you, it's not that good of a deal. Personal vehicles seem cheap because people treat them as a non-optional fixed cost, but between the cost of the vehicle itself, insurance, fuel, maintenance, and your time spent driving you can price a car service fairly high and still have it compare favourably to owning a car.
The trick will just be to convince people to start thinking about a car service the same way they currently think about their car payments or fuel costs. I'd be curious to see what a per-month subscription service would look like instead of a per-ride fee structure like taxis or ubers currently use. $300-400/mo for unlimited rides would be, for me, a better deal than owning a car. I'm curious if that could be profitable.
For my truck the numbers are different, but every time I check it turns out you can't actually use a truck for a lot of the things I use my truck for, so even though my per miles costs are something like $5/mile I don't really have a choice for those uses.
this is probably the differentiator. I bike to work about 20% of the time in the winter, 80% of the time in the summer, and don't use my car for too much else. So the cost of the car and insurance amortized out over mile driven ends up being pretty high for me.
It's roughly a $50-$70 ride for me every morning to work. That already puts me over triple my car payment. The other random trips to the store/etc would be far cheaper, thankfully, but depending on usage I can't imagine it being less than my gas costs in a month (roughly $50/m). There's still repairs/insurance/licensing/etc, but I doubt they're costing me the $700 extra that just driving to work would be costing me via taxi.
The problem is I don't live in the city, so my distances are greater, and tend to be very anti-taxi.
I actually want to take an uber to town on the weekends so I can drink/etc at dinner, but I usually don't because of how expensive it is.
Mind you, I don't think it's unfairly priced currently - but that still doesn't make it as cheap as I'd need it to be to use all the time.
That being said, owning a car is still a must if you have anything close to an outdoor lifestyle. For example:
Leaving the car at the trailhead several hundred miles from home for the whole weekend.
With Uber and Lyft, I already see my friends living an Urban lifestyle completely dumping their car. When this will become automated, another portion will likely dump their car.
But for now at least, I cannot see the outdoor crowd, or the ones living in a real rural area (think opposite of the SF hipster) dumping their car.
Nothing honors the beauty of nature like polluting it.
You can take the Long Island Railroad to Montauk.
New Jersey Transit will take you to the shore.
There's a bus that will take you to Hunter Mountain to go skiing in the Catskills.
Look up maps of the above 3 train systems. In many cases you can take a train, then bike the last couple miles, and end up in a very very isolated place.
I can also go to places that feel isolated and rural like the Jamaica Bay, the Pelham Bay Park area, or Inwood Hill Park, but are technically not because they're so adjacent to developed areas. But they do feel locally rural if you avoid a view that lets you see the city, because there is a jump discontinuity in density.
Also, you can do traditional car trips and just rent a car for the weekend. City people do this crap all the time. You really think we don't get outdoors, huh?
Demand more from your transit systems!
Personally, I live in the city and like getting out but I just use Gocar (pay by hour car rental)
(Lyft was originally just a little side project of Zimride, which was meant to link people who were already going on a trip between the same points)
Yup, that's my concern. Hypothetically a lot of things should be cheaper, but without market pressure, there's no reason for it to be cheaper.
Eg, if we pretend that an AI company releases an amazing self driving car everywhere tomorrow. Years ahead of their competition. That means the only market force they have is to price it under taxis. Perhaps they could price it at as low as possible to get more people using it, but without the market pressure it seems less likely. ... then again, I'm clearly talking out of my element here, not an educated response haha.
Personally, I'd much rather buy a car with this technology, than rent it.
https://newsroom.aaa.com/tag/cost-to-own-a-vehicle/
If you have a cheap reliable car, paltry insurance, and don't drive much you might just get down to $300-400 a month, but I think you're also underestimating your cost.
As long as the car payments aren't high and the commutes not really long, it's pretty reasonable. If your commutes are that long, a service would be outrageous. E.g., a company paid for my Uber trip from SF to Los Altos back in 2012, and it cost $125.
I'm a statistical outlier fortunately. My car with excellent insurance costs me < $100/month.
If we assume the worst case scenario where the car is worthless after the next 60k miles and I have to spend $1k repairing it, that's a total cost of 5.8 cents a mile. At 10k miles/year, that comes out to $48/month. I don't pay interest on a loan for it, but even if I did it would only be another $8/month at 4% APR.
It gets 25 MPG, at 10k miles/year and $2.50/gallon, that's another $83/month.
My insurance is $30/month. $48 + $8 + $83 + $30 = $169/month, well below $300-400.
FWIW, I actually only drive about 6k miles a year, so my actual cost is only $29 + $8 (opportunity cost) + $50 + $30 = $117/month.
I also had a car that was reliable, a 2002 CRV...well, it was reliable until we got past the 150k mark or so, then it started costing us big time to replace shit.
As others have noted, your repairs probably don't even cover basic maintenance like battery, oil changes, or fluids.
You don't have registration/sales tax in your value of the car.
Do you ever pay for parking or a car wash?
You should probably also add in any taxi or Uber/Lyft rides you already take to get a total transportation cost.
I don't think so, it's just a cheap car. It is well above state minimums (I think minimums are 25k/50k and it's 100k/300k). It includes comprehensive coverage but with a $2k deductible (mostly just useful for rental cars since $2k deductible is most of the value of the car). It has been as low as $20/month in the past.
>As others have noted, your repairs probably don't even cover basic maintenance like battery, oil changes, or fluids.
I specifically mentioned that they did.
>You don't have registration/sales tax in your value of the car.
Ah that's true, that adds about $20/month with emissions testing too.
>Do you ever pay for parking or a car wash?
Fair point, but that's less than $2/month on average.
That's crazy cheap. The cheapest I could find for state minimum in Georgia was $60 and that was answering the questions as a perfect driver.
>I specifically mentioned that they did.
But the numbers don't add up. There's no way you spent just $300 in 60,000 miles on maintenance. Oil changes alone would be $300ish. It should also be a brake job, new tires, and things like new wipers. On top of that you replaced an alternator, battery, and a belt change?
Hm maybe it's just cheaper in Arizona for some reason then. Could definitely see a significant difference in state regulations causing that.
Oh shit I forgot about tires, you're right I was missing some things. Better numbers look like this I think:
Battery: $100
Belt: $20
Alternator: $170
Coolant: $20
Windshield washer fluid: $5
Air filter: $10
Tires: $500
1x Wiper blade: $20
4x oil replacements (I replaced when oil looked discolored, not at 7k miles): $100 ($25/each)
No brake job yet, but that's probably coming soon...
Total: $845
Lack of rain/snow/salt on roads probably reduces maintenance costs by a lot in regions like this, so national averages might be higher.
So after both adjustment my numbers are probably closer to $200/month, and I'll accept that's probably an outlier and $300/month is a reasonable estimate for a cheap car.
https://www.businessinsider.com.au/googles-waymo-reduces-lid...
Sure, costs will come down, but the people who have the data (e.g Waymo and GM) seem to be banking on the taxi market, rather than selling these cars.
Kids are a major kicker, for much of the population, until cars are not only self-driving, but can also on-demand automatically reconfigure with a variety of different configurations of child restraint systems, so that you can on-demand get a vehicle configured for your number, age, and size of children.
If the safety promise of self-driving is delivered, car seats will become unnecessary.
If the total rate of accidents with self driving cars is so low that adults don't need seat belts and airbags, that might be true.
I don't think anyone projects that to be the near-term result of self-driving cars being mixed with human-driven cars on the road.
That isn't to say the service won't be useful for many even if the answer to most of those questions is "no". But urban environments are challenging even for human drivers [1], so I wouldn't be shocked if it's another decade or more before we have true, go-anywhere-at-any-time robotaxis in SF.
[1] http://www.sfexaminer.com/uber-vehicle-takes-wrong-turn-gets...
“Safe” changes on a daily, even hourly, basis. An unrelated accident changes the Safety of a particular route. A rainstorm can surface oils from the road, making it unreasonably slick for a short period of time. A crew working on the sewers can change the routing of a road onto the shoulders, which will have different grip characteristics than the road itself.
Even Arizona can see snow.
I’d sure hope so considering there’s ski resorts there (Flagstaff). Summer time highs in Flagstaff peak out in the 80s usually too.
Self driving buses will reduce the price slightly though. And importantly, people today often drive because of the "last mile" problem. Driverless taxi on the other hand, may make it much easier to connect to mass transit seamlessly. You could even connect more than once. I could get from suburban Queens to the suburbs of Philadelphia by taking a taxi to the train station, a train that is faster than cars due to traffic, then another taxi on the other end, seamlessly. Actually I can do that today, it will just be cheaper, that's all. So more people will do it.
But what if in the far future people stop owning cars and so everyone is hailing a self driving car? Perhaps then there would be enough people that you would almost always carpool? Would buses still be necessary then? I'm sure they'd still make sense in some cases but I can also see some bus stops being replaced
Even assuming 4 people to a sedan, the space needed for the engine and the safe following distance makes the density very low compared to a bus/train that is long and can have people standing packed in.
4 people to a sedan only gives you a factor of 4 improvement, which is just not enough for places like Manhattan or Tokyo. They would still need to build a dozen down from like 50.
I think the difference is important because people here frequently assume that true autonomous taxis will be here in short order. In contrast to experts who believe it is likely to be a decade or two away. E.g.: http://rodneybrooks.com/my-dated-predictions/
100% safety on 90% of possible trips is a damn good business, let the taxis take the other 10%. That is “autonomous taxis are here.” 90% safety on 100% if possible trips is how you kill people.
I agree that doing 90% of possible trips could be a good business. But it's not the same thing as true autonomous taxis. And we don't have any idea what the actual percentage of achievable trips is. 20% would still be a good business if you pick the right 20%, but it's no threat to Lyft.
In the New York Times interview a few weeks back, the CEO brought up the specific issue of store parking lots and loading groceries while the car waits, so it seems like they've thought of most corner cases.
… If it came to it, would it kill the rider or a pedestrian? What if the pedestrian is a baby in a pram? Or a dog? Can I change these settings if I care about dogs more than Alphabet's programmers? …
Arizona is a much more reliable location for these systems in large part because it’s so dry.
Listen, I don't understand why Gatorade comes in flavors other than orange, but if I was Gatorade, I wouldn't discontinue all the other flavors. Some people seem to like them for some reason.
The Darpa Grand Challenges which really kickstarted the modern self-driving car era were races.[0,1,2]
Also, there's a grassroots community around self-racing cars.[3]
[0] https://en.wikipedia.org/wiki/DARPA_Grand_Challenge [1] http://archive.darpa.mil/grandchallenge/ [2] https://www.wired.com/story/darpa-grand-urban-challenge-self...
I think the primary allure of the Jaguar brand is that dignified, beautiful, English car exemplified by the old XJ (the car from Shaun of the Dead is an XJ). It's allure is not "I'm gonna hoon in this thing and hit 0-60mph in under 4 seconds."
So they had a sporting history which like other car companies they're now resting on.
They also had a history as being the car for the self made man. The MkII was the car for bank and train robberies at the time. You'd meet a lot of Jag owners in jail...
Humans only at the beck and call of self-driving cars who cannot see well in the rain. Instead of "surge pricing" for high demand, it would be "human pricing" to pay for the "skilled" driver in the rain and snow.
ie: $1/mi during sunny days, $10/mi during snow days, instead of $7/mi every day of the year. Very interesting to consider, and honestly a bit scary thinking of so many people I've met who are using Uber to make a living. :-S
Google Maps has a ridesharing comparison function built-in; you don't have to replace incumbents to undercut them when you have the advantage, and the tech will keep improving.
It is true that if their tech can't handle rain well, they would avoid rainy cities. But even if their tech can handle it pretty well, why pick a city that has less than maximum odds for success? Going for the low-hanging fruit first could very easily just mean that they know how to prioritize, not that they can't reach the higher fruit.
Also, there are lots of other factors (like regulation and other driving conditions) that dilute the signal.
But that's very different from launching in 2 or 3 cities that don't have a lot of rain and also have a bunch of other things in common.
Depending on how they are coded I feel this will be not a big deal at all, or a huge, huge showstopper that will require rearchitecting, but not much in between.
Ice and snow is the real terror, which affects driving discontinuously and in many different ways. You can go from almost total control to almost total loss of tire grip almost instantly with black ice.
Unless, of course, they want to serve rich people first.
If they were serious about taking on Uber/Lyft soon I would expect them to be willing to make a bet on a long-life, extremely cost-focused custom vehicle. Maybe this is a stop gap to keep proving the tech until someone else (Uber?) gets impatient, licenses their tech, and spends the vehicle capital.
For a self driving taxi, I don't understand the need for a cost focused option.
It shouldn't matter too much though, drivers are very expensive compared to cars. Not paying a hypothetical Uber driver $100 a day could easily finance a supercar, let alone a Jaguar. And the autonomous car can work triple shifts.
In the unlikely event of Waymo actually ordering 20k Jags, the economies of scale they'd get would be enormous even compared with non premium Uber vehicles bought on finance at dealer prices even before the driver gets paid, and many taxi markets use premium vehicles by default.
FWIW.
Edit: ok I’m sorry I made one joke comment referencing arrested development. Sue me.
The fact that they're trying to market premium self-driving cars is interesting, because as you say there is no obvious reason to care. After all if they're all going to be driven exactly the same way and have an almost identical safety profile, what difference does it make?
It's marketing for Jaguar and Waymo. And also, Uber has premium categories, too. Its got virtue signaling potential for the users.
And, to be fair, people have aesthetic preferences. I do. If you said to me, "this cardboard box will autonomously take you anywhere you want to go without ever crashing. Or you can take this Audi S7 will do the same but there's a 5% chance you'll crash", I'd take the Audi. Personal preferences can't just be ignored, especially in the auto industry.
The same difference that makes people buy a Jaguar over a Chevy Cruze in the traditional sense... There are consumers who expect a high quality of service for their dollar.
It makes sense to have a premium brand to first introduce this.
'THIS IS THE FUTURE !' is what they want to promote.
While I think that their beetle-like test cars would be perfectly fine once the technology is everywhere, it makes sense to showcase it with premium cars.
NYC is already one of the most regulated environments for taxis in the country, maybe the most. Details down to specific models and equipment and driver behavior and pricing are specified by law and enforced by the TLC, and no driverless taxi service could comply with them as they are written today.
And if they did it using safety drivers, assuming the equipment complied with TLC regulations, they might have a hard time finding safety drivers willing to take the risk when they could just drive a normal cab. TLC licensed drivers have a lot of other requirements and liability and responsibility put upon them, and they cannot delegate that responsibility to anyone else.
As just one example, if the Arizona wreck had happened in NYC, Cooper's Law would have required the safety driver to have their license summarily suspended until the investigation is complete, then revoked permanently if found guilty of ANY charge, even a minor one like speeding 2mph.