AI can already book your hotel and order your food, but now it's coming for your ride.
In China, platforms are turning ride hailing into a single sentence.
You say what you need, for example, a smooth trip or a stop along the way, or a car that fits a child or a pet.
And then AI handles the rest.
How we order our cars is changing fast.
After that, as National Reading Week comes to a close here in China, a new report shows online literature is a cultural and economic force.
Writers are coming from all walks of life and stories are leaping into print, comics and TV.
You name it and it's there.
But as the industry goes global, it also faces challenges as it races to set new standards around AI use and its quality.
Coming to you live from our studios in Beijing, this is Roundtable.
I'm Steve.
Thank you so much for sharing your time with us today.
And for the show, I'm with Yushun and Yushan.
First up.
From ordering food to booking hotels, AI has already reshaped many parts of our daily lives.
But now it's moving into one of the most essential real-world services of all getting from one place to another.
In China, major platforms like Didi and Alibaba's QN are racing to turn ride-hailing into a one-sentence experience where users can simply say what they need.
For example, a smooth ride. or a fresh air car, meaning the air in the car is quite high quality.
You can request a trip suitable for an elderly parent or a child or a pet.
Whatever you want, you put it into the app and you let AI handle the rest.
But this is about more than just convenience.
It's about bigger questions involving personalization, digital inclusion, platform fairness is a big issue as well and whether smarter software can truly deliver in the messy realities of everyday life.
So I guess the question is is AI making mobility more human-centered or simply raising expectations that the real world may struggle to meet?
On the surface, Yushan and Yushan, good afternoon to you both.
On the surface, this looks like amazing news that is going to benefit everyone, and it is very cool and it will be a big benefit to a lot of people, but it's it's much more complicated than that.
Drivers will be affected.
There is the logistical potential problem that the platforms could be facing in terms of demand for particular types of cars.
There's a lot going on here and we're going to try to tackle it all, but first Let's start with just the big news itself.
AI can now be used for ride-hailing purposes.
Yeah, that sounds really cool when it's first announced that way, right?
Because over the past month I noticed two very similar developments that's really taken place in China's mobility market.
One would be Didi, which is one of the leading e-car hailing service provider that's been around for over a decade.
They officially launched their AI travel assistant, Xiaodi.
It is designed to handle more complex travel needs, such as trips for the elderly, families with children and maybe the need of carrying a pet large size pet dog or cat, and business receptions too.
Then we have Q1, which is under the umbrella of Alibaba Enterprise.
They introduced also a new AI ride-hailing feature that allows users to really just say one sentence.
Tell it what I'm heading from.
Say, for example, from point A to point B, what kind of car I need for that?
And then immediately, it's going to help you to get a car that's featuring all the needs that you've just announced.
Suitable to your needs.
Yeah.
Under Alibaba, right?
QN.
Sorry, what's the Chinese pronunciation one more time?
Qianwen.
Qianwen.
This was in the news recently.
I want to say a couple of months ago because they launched their AI food delivery services.
And I remember there was the whole big bubble tea explosion where if you used QN then it would give you a free bubble tea delivery, and bubble tea shops were completely and utterly overwhelmed.
Exactly.
That was the news.
And people were thrilling getting all of these orders.
And you can just ask once and or give one prompt to these chatbots saying that I want a bubble tea right now and they will pick one, maybe randomly a bubble tea for you, or you can get in and choose.
And same theory of these car-hailing platforms as well.
So you can give one prompt of your needs or demand, and then they will choose the appropriate ride for you.
Okay, so what would be?
I want to hear some different examples actually of what Maybe people are putting into the apps.
You said one sentence right.
It can be done in one sentence.
What would be an example of something that someone might put in?
So these are some stories that actually happened in March this year.
One user that used DD, and this person is located in Beijing and he just told the app that I need to go from home to Beijing South Railway Station right now and I'm traveling with an elderly person and a child and the car must have fresh air.
So that's the need that was put into the ai and within seconds, the system matched the request with several vehicles tagged with service labels such as spacious interior and smooth driving and older free.
So that's basically how the ai works.
That's pretty cool.
Yeah, to be honest, It is cool.
But the thing is that the tag of all of these cars, it's been there for a while.
After you maybe finish your ride, there is going to be a comment area and you can write comment for these car drivers or the car for the whole ride, saying that it is a spacious interior or it is like a fresh air car.
Do you click those buttons?
Yeah, of course.
Do you really?
I sometimes do, but most times I do not.
Yeah, what we're talking about here is kind of like the review process.
Of after your ride is finished.
You can give a grade to the driver and you can also give a grade to the car.
And it'll give you some different options on a screen and you can kind of click those buttons whichever is applicable to your ride.
And now they're using these comments or tags, I think, before the ride.
I think more of these comments because...
Now, previously they were just using this for limit the behavior of some of these car drivers or maybe offer telling you that we offer these kind of services, but now they're actually categorizing different tags and maybe offering you different options of these different cars.
Yeah, instead of just using it as an incentive mechanism for the drivers, now it's being used as a pool for selecting the right car for your personal need, like as if it's your personal characterized ride that the AI is arranging you.
But it's not the only thing that AI is doing nowadays for online car hailing.
Also happened in March, but this time we're traveling in Hangzhou and another user of Tianwen is telling the app to help me book a ride to LingYinSi, which is one of the of iconic travel sites.
Tourism sites in Hangzhou.
And this person said I need a ride to Lingyunsi, but I need to pick up a friend first at the entrance of Songcheng.
So that's a one sentence prompt that's put into 前问 and the system automatically understood the fact that okay, there's going to be a stopover between the location that this person is giving.
So I need to pick up someone first and then go to 另言思.
So that's just how fast and convenient it's made for us nowadays.
The AI analyzed the sentence and realized what was needed and assigned a car with those instructions, already given to the driver.
There's also a playful side to this.
I guess this is just kind of a side note to the story because I don't think this will be...
A commonly used everyday type thing.
But there was also a passenger who asked to a ride to, quote, a place to get rich overnight.
So the AI responded by suggesting some nearby lottery outlets where you could buy some lottery tickets.
That was one example of it being playful.
Then there was another request to go somewhere without work.
But that prompted a little bit more of an empathetic reply from the AI, because it encouraged the user to take a day off from.
That's nice.
So my ride-hailing app is now my therapist, in a way.
Okay, at least it explains that these kind of AI, they are trying to understand your request.
Yeah.
And they're analyzing and giving you, I think, related answers.
Mm-hmm.
As a car-hailing platform.
It's the conversational nature of AI that has made this a possibility, I suppose.
Anyway, that's just kind of a side note, a little side story to it.
But this is gaining a lot of traction among travelers, too.
Yeah, and also, I think, platforms like Tianwen.
They've been working a lot on building their volume of customers In the past few months, if not a year.
So during this year's Spring Festival holidays, for example, 130 million users here in China tried Tianwen's AI-powered shopping features.
Because, we mentioned earlier, Tianwen is under Alibaba, which also runs one of the major online shopping platforms here in China.
So 130 million users.
They tried this like shopping via AI for the first time, including more than 4 million users aged over 60.
They were building their customer base.
Basically yeah, they're attracting people in and then building trust among customers so that they already have a base of customers using more of their AI features, such as the one, starting from very recent time, for AI car hailing.
And I checked a little bit of what people are buying via AI with Tianwen.
So it seems like they are recommending more of a consumption electronic products.
Like for example, if people say, I have two cats in my home and I need a cleaning robot.
What do you recommend?
There will be a list of things based on the price and need, like how big of a home you're trying to serve, stuff like that.
And then there's also, I'm going on a mountaineering trip for this spring festival holiday.
What shoes, or what kind of sports shoes or mountaineering shoes do you recommend?
Then another list.
That's really interesting because from a business strategy point of view, i'm thinking that for them they were using that opportunity to kind of prepare customers for how to use the ai in the ride hailing service that they were obviously very aware that they were going to be rolling out within a couple of months.
It's a really interesting battle, isn't it?
The AI race among the tech giants here in China.
And you might think that on the surface and it is, I suppose on the surface it's just simple tech upgrade to the ride-hailing apps, but there's something more significant going on here, so let's talk about that.
Why now?
Why are ride-hailing platforms investing in AI powered taxi bookings?
Because these ride-hailing apps and these ride-hailing companies they were doing just fine already.
Yeah, because AI, or rather general purpose AI, they've become increasingly capable of directly handling specialized real world tasks.
So at this point, single purpose utility apps such as you know, traditional apps that just does one thing like getting you a ride via online apps they kind of need to catch up.
They need to evolve in the AI direction if they want to avoid being replaced in the first place.
Or left behind.
Exactly.
So this is one of the major push that's kind of encouraging ride-hailing industry to move away from the general one-size-fits-all services that they've already served for over a decade and towards this more refined personalized, characterized kind of service industry that we're heading into now.
Sorry, I was going to say, what makes them think this is going to work?
Because it's a little bit it's simple.
It's a simplified version of what it was before.
But at the same time, it's almost a little bit more complicated than what it was before.
If you don't know how to use it or how to talk to it.
Yeah, giving out that correct and appropriate prompt is actually one of the biggest difficulties when we're using a lot of these AIs.
So one thing is that, of course, more users are using these AI features on these platforms and, of course, booking a ride.
It can be easier because you are only now giving out one prompt and they are giving you several options.
You can choose whatever you want.
But another thing is that I think, when you think about it right, you got these options, but before you just type in where you want to go.
So that's also, I think, one step to do.
But now you need to now think about, okay, is this the cheaper option or is this...
The option with fresh air.
Is this the option with a smoother ride?
Which one I prefer?
Yeah, that's another step.
And step two, three, four.
Yeah.
Of course, this gives you, I think, a better experience when having that ride.
But another thing is that a better experience comes with higher price.
I think that is the thing comes to my mind first when I'm going to use these kind of services Because okay, there is also another option saying that I need a cheapest ride to my location, to my destination.
Well, actually, we will talk about in just a few minutes what some of the more popular prompts are.
Uh, now, but I think that's a great point.
When you were booking a ride before it involved multiple steps.
Number one, you'd have to open the app.
Okay.
Well, that's not going to be replaced.
You're still going to have to open the app.
But then you would have to enter your destination.
You'd have to choose the type of vehicle that you wanted, and that would be expected price level right.
It always gives you some options about do you want a bigger car, a smaller car, do you want a taxi?
It gives you some price differences.
That's another step.
Then you have to confirm the time or the route.
Then you have to.
If you were going to add some notes for the driver, you would add that in too, and then you would eventually complete your payment.
Now that's about, what, six to eight steps in total.
And now what they're saying is you can do all of that basically in one sentence, in one step.
Well, open the app and then one more step and then you're done.
That's saving time.
And also, according to some experts, it's called interaction efficiency, that the AI is trying to do for us nowadays.
And so, in practice, what it actually benefits for us general online car-hailing users is that it's made more personal and it's saving a lot of time for us.
So this is according to the official explanation from Didi, the platform.
If its AI ride-hailing system detects phrases like i'm not feeling well or i get carsick very easily, then their system will automatically activate service tags such as smooth driving and gas powered vehicle, because i tend to believe that for a lot of evs, no matter how big of a scale is covering in china nowadays, When they hit the brake they hit it pretty hard.
And the car stops like that.
So, yeah, for people that are easily getting carsick, such as myself, I would prefer to avoid EV.
And also there are special cases, for example, if it recognizes that the passenger is saying that oh, I'm traveling with a pregnant lady.
It prioritized tax like smooth driving and spacious interior.
So, you know, all of these small details is what matters now.
But another thing also I was thinking is that they are actually just simplifying the whole process, even the process of payment, which means that they can sliding.
So I think. you've got less chance of choosing the price that you're going to pay.
So you, of course, are getting more and better experience, but it's also a higher chance that you're going to pay more.
And another thing is that, of course, now people are liking them and these kind of sorts appeal to us.
That's also because right now it is a stage of they are trying to attract more users.
They're trying to keep more users.
The whole car-hailing app platform market is kind of saturated.
There are certain platforms that most of us are using and we are hopping on different platforms to try to compare the price.
And now they're trying to use these kind of things to actually attract more users and trying to keep them.
And that is also a way that maybe a lot of people not realizing that when you got used to using one app, it can be really hard for you to switch to another one.
Yeah, so this is, I think, a step, or a lot of these platforms are using these ways, giving out coupons, incentives to trying to attract more users.
Yeah, it is a significant change.
You know, all of the things that you said before, for example, the pregnant lady example.
Well, if the AI were to hear that, then it would recommend a spacious interior, like you said.
But then it also combines those preferences with real-time factors like traffic conditions or time of day, and vehicle location and driver status.
And what the AI does is it will filter out the options.
And this all happens very quickly, by the way.
It will filter the options from its dispatch pool and then it will send the user a short list of recommended vehicles for confirmation.
So I see that as good news, the final thing that I said there.
You're not just assigned a car with a price.
You still have options.
And there will be some price variation there, which is good for the customer.
It might add an extra step to the process, but it still allows you, the user, to make the final decision.
But if AI is now finding its way into ride hailing and the way that we are choosing our cars, I'm wondering what this is going to mean for the industry in general.
I'm thinking that it needs to adjust very quickly, because the passengers are having to adjust very quickly.
Yeah, because if we look at how e-car hailing has been around in China, I mentioned like it's been At least favored or used by people ever since 10 years ago, over 10 years ago.
It hasn't been that long.
In 2014.
I think that's when.
That's also around the time when Uber was entering China and really giving a huge boost to already existing domestic e-car healing apps and attracting.
That was also when these apps are attracting abundant of people's attention and it was really a big business for both the drivers and for the platforms.
And Yeah and back then the service mainly provided just highly standardized, if I may say, kind of service getting passengers from point A to point B, because it just started previously.
If you want a car, we kind of need to.
You know, stand by the road and stop a taxi.
That was how.
But then we can really design the route, we can choose what type of car and besides that, it's like You're lucky.
If I may say, you're lucky if you like your ride because you get what you're assigned.
And that's why I said the companies have to adjust.
Yes.
And maybe I didn't explain that well enough.
What I meant was you're now putting the preferences into the hands of the customers.
Before, they were confined to...
The screen and the options that were available on the screen.
Now those review buttons like you were talking about before.
Yushan, the car smelled nice, the driver was kind and courteous, or whatever.
As far as I know because I'm not an expert on how to use the app not 100.
That came at the end of the ride, but I didn't have any control really over what kind of car was going to be put on my list.
Now, with the AI, you're putting that power into the hands of the customers.
Right.
And that's what I meant before when I said the car companies are going to need to adjust because they need to fulfill.
If you're going to offer those requests to the customers, you need to fulfill those requests.
Yeah.
It's one of the key questions we're asking nowadays, isn't it?
Because every time when there's a new technology, We've got to make sure that this technology is really meeting what's needed out there for the general public.
Because user favorite tags and also those labels, they are quite universal if you really look into what people are asking, requesting from the AI.
And this is according to data provided by Didi, that the top personalized write requests among all the users that's been using their AI platforms are either fast and affordable.
That alone accounted for nearly 60 of the requests.
And also fresh air, meaning without odor, that's over 12 too.
And then the nearest car, meaning that the the like geographically closest one to me also features like around 10.
And then other frequent requests would include Won't cause car sickness, better quality car oh, I want a bigger car, better service, smoother ride.
So if everybody wants to get their need fulfilled, what's for the platforms on their side will have to build a large enough transport capacity in the first place.
That was going to be my next question.
Right.
That's the statistical breakdown of what people are looking for.
But can the platforms actually handle this volume of requests?
So in order for the majority of the customers to get what they want, the platforms will have to build a kind of large enough transport capacity in the first place.
And industry analyst Liang Zhenpeng argues that AI ride hailing can only work if there is sufficient transport capacity.
And in other words, that is abundant.
Supply is the real word foundation that turns what you ask is what you get, into reality.
Only if they have huge database of the review of all of these cars and car drivers they can give you the option of this car is clean enough, this car is big enough.
And here's where we're transitioning into the potential problems, right?
Because let's say, for example, that large trunk was the number one, That's not going to be number one.
Let's give a more realistic example.
Clean, fresh air in the car.
Let's say clean, fresh air in the car was the number one request from customers.
If you don't have enough vehicles to fulfill request from any given number of passengers on a particular day, then you're going to get an unhappy customer.
Wait a minute.
I already told the AI I want a car that has really clean, fresh-smelling air inside, but I didn't get that.
I'm not going to use this app again.
That's the potential problem you're looking at here.
You can make or gain a customer like that.
You can lose a customer like that.
So with these companies nowadays.
Now this is the part where we don't really have the data on really how many cars are being assigned, because this is just a new beginning for these functions to start operating as they are now.
But in the meantime we are looking at this issue as in.
Yeah, for the customers it might be unfair.
Some get what they want, some they probably won't.
And then that also brings out another fair concern, or unfair concern, about algorithm imbalance, saying that yeah, if you know those cars that features the most the best tax, that this is also a clean car, that's well attended by the driver and they're driving very smoothly.
Blah blah, blah.
Will they be the only one that gets the majority of the orders in the future?
That's looking at it from the driver's side too.
Where will this lead to if this Pandora's box is open and everybody can give requests on what kind of car they are expecting?
And yeah, for the answer of that, I kind of look into it as well.
The answer will be no, because the platform also has its own way of running the whole pool regarding how to assign the different car orders more balanced.
Yuxin and I are both DD drivers.
And Yushun just bought a new car last year.
It has a spacious trunk.
It has a beautiful interior and nice clean air.
My car is from 1989.
But before, it didn't matter because...
I would be assigned an equal number of passengers that Yushun would.
But now, nobody's going to request my old car that doesn't really smell so great anymore.
They're going to want Yushun's car.
So he's going to get more orders dispatched to him than I am, to the point where I'm not going to get any orders dispatched to me in the future.
This is the direct impact on the riders and on the drivers that.
This is their livelihood.
This is what they do to feed their families.
This is serious for them.
Well, as we just mentioned before, it's not that everybody's using the same prompt.
It's also because, as I said, a lot of people would use the prompt of I need their cheapest car, which means that, of course, the cheapest one, Which is probably my 1989.
Yeah, some of the cars can still be assigned a lot of orders, to be honest, and another thing is that I think this is, away from another perspective, motivating a lot of these drivers trying to offer better service for their customers and that is the way that you can get more orders and that is way they can improve the experience that when we are in your car, so clean your car on at least maybe you wash the.
I know what you're saying, but I think too inevitably you're going to see some drivers leaving leaving platforms.
And the problem with that is that that problem therefore gets pushed back to the platforms, because you have a reduced number.
You suddenly have a reduced number of drivers, of drivers available and a reduced number of cars available.
If the platform is a company, then their drivers are their employees.
Just like over 10 years ago, when this service all of a sudden started to operate.
They gave the drivers a lot of incentives.
Some of them can earn over like 10000, 20000 yuan per month, which is high enough salary, higher than the average salary that one can get back in 2014.
But nowadays, yes, you shouldn't.
It kind of filters out a lot of drivers' standards if they couldn't or fail to provide what's enough.
But also from the platform's point.
They hate to see that their employees are leaving them.
They don't want to to lose their employees in the first place.
So that's also why their ai or not, how tech savvy or not, their platform, their algorithms in the first place, is not really designed to assign things in balance, to just give orders to the ones that have the better reviews.
And also for their customer side.
If everybody wants the same best quality service, then that probably would mean that There will be longer waiting time for the customers.
And you don't want that either.
Yeah.
Yeah, exactly.
So the reduced number of drivers in the future, if that happens, get that problem goes back to the company and then that problem goes back to the customer or the
The request simply cannot be fulfilled.
And in that case, Didi, well, they have a solution in place for that.
So if I request a car that has perfect air.
For example, the AI assistant Xiaodi you said was the name.
It's going to rank my requests by priority and And then satisfy the most important needs first.
So if I said fresh air is number one and closest car is number two.
Well, if it can't meet the fresh air priority for me, it will go to the second priority, the closest car available.
So I guess that's their way of dealing with the potential issues problem, where not every single request from a driver is going to be fulfilled.
But but, but.
If every single request from the driver can't be fulfilled, well then what's the point?
What's the point of this?
Do you see what I mean?
But I don't see how the customers are going to end up by saying like, oh, well, maybe next time.
Right.
If they're going to be paying a premium.
By the way, I don't have pricing plans in front of me.
But if they're going to be paying a premium for this service, I would think you better meet those needs or you could be in trouble.
Or I can see, is that a lot of these platforms.
They're trying to offer you the more comprehensive service.
Even I remember that some of them, a lot of them even like food delivery platforms, they're introducing these membership system that you can even link to other platforms, getting a, like video streaming platforms, vip through that.
Single vip yeah, single membership.
Yeah, this um dd platforms also has that, so they are trying to get their apps into a super app that we were talking about.
You can do everything in that one single app, so that you spend more time on this app, you spend more money on this app and they have more cash flow, or the whole platforms can be more thrived.
And the way I see it is that this is not going to be an overnight kind of transition.
They have this AI function, let's call it, already out there.
But in the meantime, their traditional car hailing service is still there.
It's not like the traditional ones that kind of just get standardized, as we mentioned earlier, is like erased all of a sudden.
So if we go both ways, like Yushun mentioned, along with all the other lifestyle ecosystem they're trying to build within their app or within their platform, then Slowly it may actually become the main function one day, many years later, that people only get AI car hailing.
But nowadays, when it's out of service, it's just simply going to tell you oh, I'm overcrowded right now.
Please try again later, like they do with my milk tea orders previously.
Yeah, right.
Yeah.
It is incredible.
There's no question about that.
I mean, imagine that you open up your app and you say, oh hey AI, assistant my wife and kids and my parents and I.
We're going to be traveling to the Forbidden City for the weekend.
So can you help me find a hotel nearby there that's in this specific price range and within walking distance?
Then the AI assistant can help you find the hotel, can give you the perfect car based on what you said.
I mean, it's just so much more convenient because previously you would have to go from app to app to app to app to book everything that you wanted to do.
And now you can do that all in one place.
And yeah, there might be some wrinkles in the road, some bumps in the road along the way.
It's only going to get better though.
It is very cool indeed.
