Q3 2024 Cheetah Mobile Inc Earnings Call
Good day and welcome to the Cheetah Mobile third quarter 2024 earnings Conference call.
All participants will be in listen only mode.
You need assistance. Please several conference specialist by pressing the star followed by zero.
After todays presentation, there will be an opportunity to ask questions.
So asking question you May Press Star then one on your telephone keypad.
So let's all your question. Please press Star then two.
Please note today's event is being recorded.
Helen Xu: I would now like to turn the conference over to Helen Xu.
Helen Xu: Cheetah mobile please go ahead.
Helen Xu: Thank you operator, welcome to Cheetah Mobile's third quarter 2024 earnings conference call with US today, our company's chairman and CEO, Mr. Fu Sheng, and our director and CFO, Mr. Thomas Kim following managements prepared remarks, well will come back.
Speaker Change: The Q&A section. Please note that the C O create what is presented.
Speaker Change: Before we begin I refer you to the Safe Harbor statement.
Speaker Change: This release, which also applies to our earnings conference call today I will make forward looking statements at this time I will now turn the call over to our CEO. Mr. Fu Sheng. Please go ahead with them.
Speaker Change: Hello, everyone.
Fu Sheng: Thank you for joining us today.
Fu Sheng: Once again achieved accelerated revenue growth in Q3.
Fu Sheng: Led by our service robotics and Internet businesses.
Fu Sheng: This consistent growth results from our strategies to expand the use cases of our wheeled servicer, Bobby and expand into overseas markets.
Fu Sheng: As well as the resilience of our legacy <unk> business.
Fu Sheng: Industry demand for service robots continues to rise.
Fu Sheng: Specially in the overseas markets.
Fu Sheng: And then restaurants hotels factories and offices.
Speaker Change: It isn't as strong as our usual robots more often to help their staff and improve efficiency.
Fu Sheng: In the past weeks.
Fu Sheng: Many customers and partners in Europe and associations Asia in.
Fu Sheng: In fact, I am still in Europe today.
Speaker Change: Jim with our local partners to further strengthen our presence there.
Speaker Change: Building a strong local distribution network is very important for our global expansion.
Speaker Change: Because it will set us apart from our peers.
Speaker Change: That's why I have spent significant effort on this initiative.
Speaker Change: My conversations with local partners.
I learned that a robot to helping them solve labor shortages.
Customers in Europe should that you didn't cheat us robots have reduced employee absences and turnover.
Speaker Change: Meanwhile, some Japanese customers told us that robots are much more reliable and other players offerings and are switching to our products in September we launched a new robot for factory and fulfillment center use.
Speaker Change: This robots can autonomously delivering goods to move low payload items from Charles its warehouses to assembly lines.
Speaker Change: We are currently optimizing the product to better meet the needs of customers in overseas markets.
Speaker Change: This highlights the importance of receiving feedback and input from local partners.
Speaker Change: They believe this product will become an important part of our service robotics business in 2025.
Speaker Change: While the robotics industry is still in its early stages it.
Speaker Change: It will be a trillion dollar market, Robert shall becoming essential hoppers for humans.
Speaker Change: It will happen in developed markets first.
Speaker Change: And L. Ams will speed up this growth by enabling robots to do more tasks.
Speaker Change: And making them easier to deploy than ever before.
Our conversations with investors.
Speaker Change: We've noticed a lot of interest in how L. Ams are making our robots smarter and driving steady revenue growth.
Speaker Change: Today.
Speaker Change: I will share what we've achieved so far with no Ms in our products and what's coming next.
Speaker Change: Yes.
We are using <unk> to improve the way our service robots interact through voice, thanks to our strong far field voice recognition.
Our robots already here user as well.
Speaker Change: Now it's L EMS.
Speaker Change: Understand users' questions more clearly and respond better.
Speaker Change: Making the overall experience much smoother.
Speaker Change: For instance at restaurants, Irobot don't just the labor food during the busy hours.
Speaker Change: It can also help attract customers.
Speaker Change: We're seeing the return on investment for our restaurant owners by taking on more roads.
Speaker Change: Because L M breakdown language bearish.
Speaker Change: Our expanding these voice enabled robots to international markets.
Speaker Change: We are also working on agent noise.
Speaker Change: System that lets customers set up costs for our robots using voice calls.
Speaker Change: For example.
Speaker Change: You can tell a robot to check each table at two P. M to see if anyone wants to order more food before the kitchen closes at 230 P. M.
Speaker Change: The rollout will go to each table skip the ones without customers.
Speaker Change: And even allow people to place orders if the ordering system is late.
Speaker Change: L L M D.
Speaker Change: Just kind of finishing out it would be nearly impossible or would require writing a lot of complicated coat side.
Speaker Change: We are using multi modality models to improve our robots indoor autonomous driving.
Speaker Change: One area that we are working is to enable our robots to map out the large factory as they move and look around.
Speaker Change: Once the map is really.
Speaker Change: R&D Jean Marc key locations.
Speaker Change: Robots kanban and navigate the factory on their own to deliver goods.
Speaker Change: Actually all of US wanted to change these key locations.
Speaker Change: They can easily update the map whenever needed.
Speaker Change: Based on our initial testing.
Speaker Change: Thanks to L Elms.
It takes to set up our robots can dramatically reduced from about two days to just two hours.
Speaker Change: Already in using vision based autonomous driving technology in some cases and plan to expand it further.
Speaker Change: Instance, armor.
Speaker Change: Our robots using vision based economist driving technology to avoid tibor and obstacles and understand their surroundings.
Speaker Change: Over time, we aim to achieve an ACH and navigation system.
Speaker Change: This will allow our robots to handle more complex environments entirely on their own.
Third we are adding robotic arms to a robot to help them to specific jobs.
Speaker Change: Some of these arms can press buttons.
Speaker Change: Which is useful for delivering things between floors in particular in the overseas markets.
Speaker Change: With business owners are reluctant to adjust the elevator access control systems due to security concerns and.
Speaker Change: Others can pick up in store items, which is great for use in factories.
Speaker Change: Arms are powered by on device multi modality models, making it easier to automate routine tasks.
Speaker Change: When it comes to L.
We use advanced models through API calls to support some of the features we have discussed at the <unk>.
Speaker Change: Same time, we are also developing our own models.
Speaker Change: In November we launched an AI by 7 billion mixture of experts model.
Speaker Change: Covering many languages, including Chinese English Korean and Japanese.
Speaker Change: Made it open source and use it to power our own us.
Speaker Change: So, especially the agent OAS features.
Speaker Change: Additionally, we have trained smaller on device models to support indoor autonomous driving and control robotic arms.
Turning to L. M based applications, where you recently introduced areas and I based data service platform.
Speaker Change: So assisting enterprises in data and building pumps for their L. M based application Eric.
Speaker Change: <unk> was built on top of our insights into developing L M and L. M based apps so.
Speaker Change: So far we have received positive customer feedback on our L. M based application.
And we will continue to enrich our portfolio.
Speaker Change: Goal is to offer relatively standardized the SaaS products, allowing businesses to use L. M is gain efficiency.
Speaker Change: Before handing the call over to Thomas for the financial highlights.
To stress this cheetah.
Cheetah mobile is in a good position to tap into the growing market for service robots and L. M based apps.
Speaker Change: Years of experience on the PC and mobile phone area as.
Speaker Change: As well as expanding into international markets and.
Speaker Change: And we have strong L M expertise.
We have already made solid progress in growing our revenue and cutting losses.
Speaker Change: This is just the beginning of Cheetah mobile its turnaround.
Speaker Change: Thank you Vishal Hello, everybody on the call. Please note that unless stated otherwise all money amounts are in RMB terms.
Speaker Change: In Q3 2020 for our financial results demonstrated solid execution on two strategic objective number one accelerating revenue growth driven by both the sales of robots and the legacy Internet businesses.
Speaker Change: And number two enhancing our operational efficiency to reduce our operating losses.
Speaker Change: Initially.
Speaker Change: In the third quarter of 2024 hour total revenues increased 16, 6% year over year.
Speaker Change: Marking the second consecutive quarter accelerating revenue growth.
Speaker Change: Compared to 11, 6% in Q1.
Speaker Change: 12, 3% in Q2.
Speaker Change: Our view based service robots continue to be a key driver of growth. Additionally, our legacy international business remains resilient.
Speaker Change: <unk> solid revenue growth and margin expansion.
Speaker Change: On profitability, we made further progress.
Speaker Change: non-GAAP gross profit rose, 14% year over year and.
Speaker Change: 7% quarter over quarter.
Speaker Change: $131 million with non-GAAP gross margin.
Speaker Change: Spending 268% in the third quarter from 65% in Q2 and 63% in Q1.
Speaker Change: non-GAAP operating loss was $61 million in the quarter.
From $63 million in Q2, and 66 million in Q1.
Speaker Change: We continue to focus on managing costs and expenses, notably we consolidated the teams of Cheetah and Beijing, R&R streamlining stop and services with overlapping function.
Speaker Change: For example, in Q3 be reduced at least cost and cloud costs professional and legal fees and certain labor cost related to G&A and operations.
Speaker Change: We are also defensively investing in AI using AI to enhance our service robotics business.
Speaker Change: Example, our non-GAAP R&D expenses increased 25% quarter over quarter in Q3, and now about 60% of our revenues are invested in R&D.
Speaker Change: As <unk> has said in the past that when we focus on developing products that generate immediate revenue and profit.
Speaker Change: Remain attentive to the latest technology called advancements.
Speaker Change: Another highlight is the continued strength of our legacy Internet business, which grew 26% year over year and 18% quarter over quarter.
Speaker Change: Operating margin excluding share based compensation expense for this segment improved to 10% from 6% last year.
Speaker Change: As of September 30th.
Speaker Change: <unk> 24, we maintain a strong balance sheet with cash and cash equivalents.
Speaker Change: 1 billion 800.
Speaker Change: And $31 million or U S dollar $280 million.
Speaker Change: Long term investments.
Speaker Change: About $886 million or U S dollar $126 million.
Speaker Change: In closing, we are making solid progress in the expanding revenues and narrowing losses, we are confident in our investments in yet because we see the significant market potential.
Speaker Change: Integrating.
Into our surveys robotics business.
Speaker Change: At the same time, we remain disciplined in reducing losses and driving efficiency in our operations.
Speaker Change: Thank you this concludes our prepared remarks.
Speaker Change: Operator, we are now ready for the Q&A session.
Speaker Change: Thank you.
Speaker Change: Okay.
Ladies and gentlemen.
Speaker Change: Sandbox or the English translation of the question and answer session.
Speaker Change: Please note there may be a delay before the term place to begin and then between question.
Speaker Change: Please do not disconnect.
Speaker Change: Okay.
Speaker Change: Okay.
Speaker Change: My question focuses on cheaters robot business.
Speaker Change: May I ask for the year 2025, what specific goal has been set for the robot business in terms of shipment volume revenue growth and revenue proportion.
Speaker Change: Yeah.
Speaker Change: Okay.
Speaker Change: Yes.
Let me answer these questions.
Speaker Change: Well for 2025, some of our specific importing goals should be said to still be in the process of formulation.
Speaker Change: I have also recently visited many channels overseas.
I think the overall general direction can be determined like this.
Speaker Change: First our base income will definitely increase in the next two five years and the proportion in the entire revenue scale of Cheetah should also increase.
Speaker Change: The specific one should be said that we will combine this wave that is by Thursday, we have summarized a lot of experiences in these overseas markets and then make certain specific assumptions.
Speaker Change: And we think of doing the setting of specific goals in this way.
Speaker Change: Then our idea is that because the robot industry, it's very hot in the capital markets today, but in fact, it's growth that is the growth in the short term I don't think it will be as highly explosive as the Internet was back then right I also have to tell everyone.
Speaker Change: One that it should be a continuous and gradually accelerating process because today for robots in the combination of software and hardware sorry. This is a bit in.
Speaker Change: In the combination of business than the physical implementation and the aspect of channel construction. The investment is much larger than that on the internet is much larger than the exponential growth Baxter.
Speaker Change: So I think for the robot business, we will look more at a long term goal. We may for example be able to make the commission income accounts for more than half of the entire company within three years and including that we can be among the top few globally at least one of them.
Speaker Change: The top three such silver system providers.
Speaker Change: Since our peak gold.
Speaker Change: And for the detailed goals, we still need some these cautions and careful deductions.
Speaker Change: The cash cow business of Cheetah that in the Internet business performed very well this quarter with both revenue and profit margin remaining stable.
Speaker Change: May I ask how should we predict the revenue and profit margin trends of the Internet business in 2025.
Speaker Change: Will it show a steady growth trend or gradually slow down in decline.
Speaker Change: I would like to raise a few questions about global training.
Speaker Change: In the process of global training, how do we overcome the problem of scarce data.
Speaker Change: What facilitating wrong Cam large models play in training data in fact, Ciena robots have deployed many robots globally.
Speaker Change: So in the future is impossible for us to continuously train cheetah robots in a data driven manner to continuously enhance their intelligence level.
Speaker Change: Furthermore, in terms of the training methods of robots, how far are we from achieving the goal of robot self learning and training by watching videos of human performing tasks.
Speaker Change: Compared to leading robot companies overseas, how big is the gap in this aspect domestically.
Speaker Change: Sure.
Speaker Change: Yeah.
Speaker Change: Okay.
Speaker Change: Okay.
Speaker Change: Yes.
Speaker Change: Yeah.
Speaker Change: Okay.
Speaker Change: [noise].
Speaker Change: Well the first.
Speaker Change: One is a rather complex technical question you asked.
Speaker Change: I'll try to use my understanding and that of our company to simply help you analyze and explain it.
Speaker Change: Regarding the data requirements for robot training it can be divided into several aspects. One aspect is that we divide the robot into several components. One component is navigation, which is equivalent to a small endorsed that what's driving this.
Speaker Change: This matter because the environment is relatively limited within an indoor space has basically been solved through some engineering technologies based on certain models and sales before.
Speaker Change: However, due to the addition of underlying data last time indoor navigation of robots will become how to say it that.
Speaker Change: That is to say its implementation will become more real time and its reliance on sensors will become smaller.
Speaker Change: This is what we are currently doing.
Speaker Change: Right one of the things we are doing recently is to promote the indoor navigation of our robots from mathematics and chemistry.
Speaker Change: Our next two robots can also be equipped with a higher level of chip purely achieving indoor navigation through vision. This is also being gradually advanced you can also compare it.
Speaker Change: Look at today's new energy vehicles in.
Speaker Change: In the end it was tesla's urban self driving that made significant progress in adding liners and various radars and multi modality actually seem to be less important now the SSP.
One important reason is the emergence of this transformer and the mechanism of such large model.
Speaker Change: This mechanism as the underlying implementation for end to end processing can handle many things. So this is one aspect.
Speaker Change: Regarding this aspect of data just as you said, we have deployed many robots and they have been running in various scenarios before this can actually achieve a considerable amount of data and the road conditions. It faces are not as complex as films on highway.
Speaker Change: Nor are the speed requirements as high.
Speaker Change: In this part we think it's okay.
Speaker Change: <unk> is not a major demand.
Speaker Change: Especially for a company like us that already has many robots on the ground providing services every day.
Speaker Change: Are we seeing that in this part of indoor navigation, we don't have any.
Speaker Change: Really the second one might be a concept of self learning skills that is currently quite popular.
Speaker Change: For this I think it is still more theoretical at present.
Speaker Change: The computing performance and exactly what kind of new driving force. It is actually don't have a particularly clear definition right.
I'd say its circular and some say it's deck.
Speaker Change: Human like or dual closed loop.
Speaker Change: The data in this aspect is indeed relatively scarce because previously.
Speaker Change: Including the robotic arms in the factory to mention were constructed based on not the data system as the core but the cold channels for automation at the core.
Speaker Change: We think our approach is to take it step by step.
Speaker Change: I have constantly expressed a viewpoint on many occasions that I believe that for humanoid robots. There is still a long way to go before they truly land and become commercialized products right, it's unlikely to truly achieve commercialization without five or 10 years.
Speaker Change: You can see that many of their demonstration effects are good but for them to truly become commercialized products. There is still a long way to go so it's more about a pragmatic approach. So we will combine our scenarios for example, we start by completing.
Some simple tasks in the interaction between some robotic arm and the real world.
Speaker Change: I won't go into the specifics of this because each related to our technical group.
Speaker Change: Our idea is not to come up with a perfect product that can do everything and solve one universal problem.
Speaker Change: Our idea is instead, because we have scenarios on the ground today, we combined the scenarios themselves more to complete the continuous collection and training updates data. We think these require some time.
Speaker Change: It may not be as optimistic as we mined thing today in terms of investment, but I think we will complete one scenario after another step by step.
You can also see that in some foreign startups the tasks. They complete with venture capital are very simple.
Speaker Change: But I think this way easier to implement if they come up with something like cooking and making meals right away is basically a laboratory product because there are too many constraints in performance, making it very difficult.
Speaker Change: For the second question, how far are we from achieving the goal of robots completing tasks by watching human videos for improvement.
Well, it's still quite far.
Most of what we see now our demonstration videos.
Speaker Change: I can give you an example.
Speaker Change: Like that time it was all over a moment.
Speaker Change: What did you do and then in wood alone anyway, but its success rate is very low.
Speaker Change: In the papers published it was 70% or something.
Speaker Change: Of course, there will be progress and this 70% is still in a specific situation.
Speaker Change: For example on the desktop is not the entire desktop, but a designated area.
Speaker Change: So it's not as urgent as we thought consider autonomous driving.
Speaker Change: Many teams have been working on it since 2016 and 2017, it's been eight years now in a two dimensional world surface situation today, No autonomous driving company has achieved the al four level right. This autonomous driving and adds up.
Speaker Change: Beginning everyone was very optimistic thinking that once recognition was achieved autonomous driving would be possible right.
Speaker Change: But today Tesla also announced that in 2026, it will have robo taxi landing, including such cars.
Speaker Change: I think the time for robots to watch humans perform tasks for self learning is not more optimistic than this autonomous driving because it's more of a three dimensional mechanical system with more mechanical structures involved.
Speaker Change: This is our judgment on the major technological trend, but as for how big the gap is between domestic mechanical companies and foreign ones I, frankly think it's not significant.
Speaker Change: Because today with large models, including after they are developed domestic updates are also very fast because the underlying algorithms. Even those that can be shared that is at the AI level. Once the algorithm itself achieves a breakthrough that <unk>.
Speaker Change: <unk> for everyone to learn is not high the real difficulty lies in how to engineer. This algorithm how to use more data for training and how to train it more efficiently.
Speaker Change: In fact Chinese teams have an advantage in this regard in the country.
Speaker Change: I believe there is no gap.
Speaker Change: In doing this kind of large scale data engineering.
Speaker Change: So today I don't think there is a big gap.
Speaker Change: This is about the training aspect of the existing methods.
Speaker Change: Including everyone using some domestic University model a P P.
Speaker Change: In fact, the productive Asian and other aspects are actually quite good.
Maybe if the gap really exist it might be some new paths. For example, if there is a new method emerging I think there will be some gap in the country is very difficult to come up with some particularly innovative methods, but once the surgeon method emerging the speed of.
Speaker Change: Mastic follow up is very fast there is not much gap. This is my view.
Speaker Change: Personally.
Speaker Change: Thanks.
Speaker Change: The beginning of this year the company has been continuously reducing losses every quarter.
Speaker Change: I ask what the subsequent pace and specific plans for reducing losses are.
Speaker Change: Is there a clear timetable for achieving profitability.
Speaker Change: Reducing losses is definitely.
Speaker Change: Our top priority so.
Speaker Change: So today, all enterprises are shouting about cost reduction and efficiency improvement.
Speaker Change: For us we have surely achieved a certain scale of loss reduction this year.
Speaker Change: But I want to say that for this loss reduction because we participated in some research and development and training of large language models, our pace has slowed down a little bit.
Speaker Change: After we review the Emmy. This round, we will put more of our efforts into the development of agents and the implementation related to the intelligence of robots.
Speaker Change: Such R&D costs will be reduced significantly compared to the large language models in the past 10 years with.
Speaker Change: We definitely have a certain loss reduction plan internally any timetable for profitability well I think due to the significant changes in the market and technology, we may not be able to disclose this very clearly externally you know because because now I think that this way.
Speaker Change: <unk> just like the France question earlier about some progress in training because we have seen some business opportunity and we also see that due to the support of large language model towards service robot, whether it's their underlying planning ability tax decision, making ability including their interaction.
Speaker Change: Ability, including their interaction bill of it and you will expand their satisfaction in various markets and expand the market and expand the market in various markets and expand the market. So we also need to maintain a certain flexibility if such an opportunity arises we may have to make some more <unk>.
Speaker Change: That means in R&D.
Speaker Change: Of course overall, we will definitely aimed to make the company profitable and creates value for shareholders.
Speaker Change: This major goal is beyond doubt.
Speaker Change: How do you view, a gen take AI or AI agent how does it.
How does this technology.
Speaker Change: For example, we have seen that recently the AI agent of G pool has been able to independently help users search and place orders on BD, Max and Matewan Ken.
Speaker Change: Can AI agent to accelerate the application of large models in the C and D.
Speaker Change: How should we consider the value distribution in this.
Speaker Change: Well, okay. Thank you.
Speaker Change: This question is quite professional.
Speaker Change: I think the popularity of these terms, which has become especially popular recently essentially four in AI agent. It is more like some traditional types of software.
Speaker Change: What was it called before.
Speaker Change: That kind of software essentially emerged because the model's ability has not reached assertion level that kind of appreciation and smaller foundation and then we need to use a part of human thinking change and a part of human planning to guide the large model.
Speaker Change: Just like I wrote the first 45 centers since like beds and after that the effect, including what you saw with the so called points of GP I think it might be a paradigm shift to a new software.
Speaker Change: That is previously when we wanted to write software for our library, we have to rely entirely on a large amount of cold and logic to complete it.
Speaker Change: But today a lot of cold logic is handled by large language model than the reduction in the R&D cost of this kind of software on this end and the improvement of its user experience at every step.
Speaker Change: It is indeed, a great opportunity.
Speaker Change: I have to say that we our selves also have.
Speaker Change: We have seen that our internet business also has some attempts in this regard.
Speaker Change: But I want to say that for this matter of an AI agent, whether it is able to complete these queries or place orders.
Speaker Change: To achieve the high level of stability and satisfaction of traditional systems. What does that mean for example, if I give it any destruction can it definitely do what I want to search for it.
Speaker Change: But this is a very crucial point for the landing of large models on the sea end or beyond today.
Speaker Change: This point is not I really mentioned in the industry, but we discovered this problem. When we were doing age ourselves. For example, if you use traditional cold tool implemented when you select a point to place an order because humans are very precise in their operation.
Speaker Change: Basically your operation is 100% similar right sometime.
Sometimes it might fail.
Speaker Change: Sometimes it might not meet your expectation and sometimes it might give you errors.
Speaker Change: The large models don't know what they don't know they have hallucinations.
Speaker Change: So this point is that for an AI agent, who truly land a lot of effort needs to be invested in this aspect.
Speaker Change: Coming back to your question can it accelerates the lending of large model applications on the C. N E S.
Speaker Change: <unk> can especially on many C N. The apps today some apps have already started.
Speaker Change: They have begun to use. These for example, I can give a few examples like image translation tight and some educational types. This has clearly started yet.
Speaker Change: <unk> for the value of distribution I think it will still bring a wave of practice for application manufacturers.
Speaker Change: So this is my view.
Speaker Change: Personally I'm, not particularly optimistic about the lending on the sea and in China as a startup company because the capabilities of large domestic manufacturers. In this aspect are extremely strong and for them. It will also be very very fast. So if you make a small innovative.
Speaker Change: <unk> in an area that others are familiar with it should be quickly follow up thank you.
Speaker Change: I'd like to ask a few questions about the application of large models in the enterprise level. We have observed that Cheetah has conducted many explorations in the application of large models this year covering areas.
Speaker Change: Such as by training sales management and data services.
Speaker Change: How willing are enterprise customers to pay for the application of large models at present.
In the office scenario is the application of large models still mainly limited to specific working scenarios with a higher error tolerance rate with the advancement of large model technology, especially the emergence of AI agent can this effectively enhance and improve the hallucination problem.
Speaker Change: One of large model, thus, enabling the application of large models to completely or partially replace manual work.
Speaker Change: Okay.
Speaker Change: Yes. So your question itself is very revealing I think you have covered a basically every point.
Speaker Change: I think the willingness of Andrew price customers to pay for the application of large models completely depends on how much input output ratio based application can bring to them and there is one point that needs attention that is these input output ratio has to be high.
Speaker Change: <unk> relatively high because essentially it means theyre reengineering of many internal processes of the enterprise and the redefinition of some positions.
Speaker Change: For an enterprise is there is not a high enough value they are not willing to promote it.
Speaker Change: Currently it seems that the enthusiasm of enterprise customers to pay for the application of large model how to say it it is becoming more and more rational.
Speaker Change: What I know is that last year. Many enterprises spent a lot of money on the authorization or privatization authorization of a large model. This.
Speaker Change: This year it is obvious that they no longer paid for these or very little instead, they ask more about what kind of directly usable thing you can provide me well.
Speaker Change: Well and your question about whether it is indeed in specific scenarios with a relatively high error tolerance range for the current large models for example in training and some summaries of sale because even if it is slightly in accurate it is still okay.
Speaker Change: Right body, you roughly see a general idea or that for something like training. If it can achieve an accuracy rate of more than 90% a large number of people can accept it manually but in some aspects such as data insight I think the entire industry is still exploring.
Speaker Change: Together.
Speaker Change: Yes, you can effectively has an improved staff hallucination problem of large model.
Speaker Change: Because he uses traditional codes or task planning to confine the ability of the large model in a particularly vertical environment.
Speaker Change: At this time when the scene of the large model is particularly vertical the probability of its era will decrease, especially in this wave the current ability of the large model has reached this level.
Speaker Change: When everyone is investing in this therefore defining the application of these large model in the B and we'll definitely gradually achieve.
Speaker Change: More and more replacement.
Speaker Change: Here just in combination with the next question I also gave an example.
Today, if we do the application of large models. If it is a fun application theoretically it is difficult to satisfy customers enough.
Speaker Change: This is after our exploration for such a long time.
Speaker Change: Let me give an example for example, you don't know if he bought the latest iPhone right.
Speaker Change: I also specially bought one that can use the overseas version to try you will find that what it can truly reflect is very legal.
Speaker Change: The experience is not that long.
Speaker Change: And including what Microsoft pushed I think both of these applications define them too broadly because they are large companies.
Speaker Change: Zinc on this point, but anyway, they have enough business class and they are slowly moving in this direction.
Speaker Change: But as an enterprise like us, we must do a very clear and vertical application.
Speaker Change: And our idea is to penetrate one point then make the experiment replicable and then move to the next point.
Speaker Change: So basically these are some of my understanding.
Speaker Change: Thank you.
Recently, many people have discussed a lot about the slowdown of the scaling law.
What's your view on that.
Speaker Change: What are the impacts of the slowdown of the scaling law on the development of the large model application industry.
Speaker Change: You really don't need to worry about it there are still some disputes someday.
Speaker Change: One point is that whether the key wide itself is slowing down or not is unknown.
Speaker Change: Or at least recently, especially in the past one or two months everyone has been talking about the insufficiency of data right because basically on the internet. The good and precise data that can be used for large model training approximately well in the industry.
Speaker Change: This is not certain but what I heard is probably around 20 to 30.
Speaker Change: It's basically about this amount of data because although there is some data its quality is not high and it might make the model less usable right, including now in our industry everyone is keeping an eye on it for example, GPT has not been.
Speaker Change: <unk> released yet right.
Speaker Change: Can see that this time in 12 days, it's basically the enhancement of the productive bashan of the original model capabilities in the past.
Speaker Change: What does this mean.
Speaker Change: It might imply that in a certain sense today top level models globally is not likely that the model capabilities will increase easily for a period of time.
Speaker Change: This has been at least four or five months right.
Speaker Change: Before then we saw G. P. T 3.0, GPT three five and then two 4.0 to four.
Speaker Change: Each step was quite fast before but now it's been a long time.
Speaker Change: So we think that at least in reality today, the growth and the expansion of the capabilities of the top large models in this industry.
Speaker Change: We don't know whether the scaling law is slowing down or not but this is definitely slowing down but this is a good thing for startups themselves, especially for companies like us that do application because when the model capabilities. We are developing rapidly before many things here.
Speaker Change: Did really when he was 3.8, let's not talk about it.
Speaker Change: When my New model came out it burden.
Speaker Change: There were indeed, some such projects at that time.
Speaker Change: After they finished when the model came out they added this capability and then it didn't sell but today due to the slowdown in the growth of the top model capabilities, everyone is thinking about how to better utilize these capabilities with agents.
Speaker Change: This is a major idea for doing application. So therefore, we think this is beneficial to US. We can also be at ease right. We don't participate in the model selection competition anymore. We just do a good job with the application itself because we end.
Speaker Change: Turn to the gang relatively early today, we still spent a lot of effort on doing some scenarios than for US. This scenario is the best way to Polish it for a lifetime.
Speaker Change: Because in the end, it's Athens steel needs to be combined with customers and the market to know how satisfactory. This thing is.
Speaker Change: So I think it's a good thing for companies like ours.
Speaker Change: As for what this means for the industry I don't have such a high perspective.
Speaker Change: Well thank you.
We understand that president foods attitude towards robots development is extremely pragmatic focusing our way.
Speaker Change: Those robots that can achieve large scale applications at the current stage rather than in the field of by Tito robots.
Speaker Change: However, we have also noticed that cheetah robots are attempting to integrate robotic arms and launch embodied intelligent robots products as defined by Cheetah I would like to ask how you view the changes in the robot industry trend over the past three months for example, what <unk>.
Speaker Change: Loading effects two large model have on the implementation of robots, how do you view the future competitive landscape of the wallboard industry.
Speaker Change: The changes over the past few months, how do I feel.
Speaker Change: In the robot industry. The concept of large model has become increasingly popular.
Speaker Change: Then due to the addition of Tesla's optimist, the humanoid robot has become quite popular several times.
Speaker Change: But I still adhere to my viewpoint robots are too broad a term so when people mention it they always have the fantasy using humans as the prototypes to create a perfect product that can solve all problems. In fact I think this path is at least very.
Speaker Change: Very difficult.
Speaker Change: We can look back.
When it comes to autonomous driving in the past initially there were two routes competing one route was led by Google's way malt, which pursued having won a highly functional vehicle with many sensors expensive lidar and very perfect algorithms to achieve.
Speaker Change: This individual was quite strong to achieve stable autonomous driving.
Speaker Change: The other route was tesla's approach at that time, which was that I didn't think I could definitely achieve it so I would start with visual sensors.
Speaker Change: Continuously add sensors eventually it was found that.
Speaker Change: When training.
Speaker Change: Up to today, we can basically say right. The second one is that with the in depth exploration of scenarios and continuous experimentation with data. The current technological changes are better than the initial perfect assumption of having the highest quality.
Speaker Change: <unk> engineers and the most advanced sensors.
Speaker Change: Currently it seems that this approach is much better.
Speaker Change: So I think it should be the same for robot it is.
Speaker Change: Unlikely to have a situation, where I am different and can complete everything.
Speaker Change: Just like humans, they still need to drive a car or use a trolley or some tools to assist to achieve more practical application.
Speaker Change: Right. So today I think although the concept of robots was very popular recently there is still a considerable gap between the concept and practical implementation rigor.
Regarding the hallucination problem mentioned in large model in fact, it is like this is a fundamental algorithm issue it has such problems.
Speaker Change: Times in language the answers are somewhat acceptable, but when it comes to actually pro forming an action or interacting with objects there cannot be any errors.
Speaker Change: And one such errors occur it will make it very difficult for the machine to be implemented right. For example, in a restaurant or in some reception scenario.
Speaker Change: You will find that when there is no one supervising it it has to operate for over 10 hours a day hundreds of times, a year and cannot make mistakes in thousands of scenarios.
Speaker Change: Little mistake will affect your customers' confidence and whether they will recommend you to others.
Speaker Change: So I think large models are definitely helpful for the general direction of robots, but when it comes to actual implementation it should be advanced to step by step in combination with scenario.
Speaker Change: <unk>, which is also why I am not very optimistic about humanoid robots.
I think for the future of pattern of this this question is very broad.
Well I believe that eventually it might be like the example, I gave more pragmatic robot manufacturers will emerge continuously.
Speaker Change: The final competitor that Polish it within a product will win.
Speaker Change: I am not optimistic about proposing a very big concept today, and then moving towards a loan lending operation.
Speaker Change: This I don't think it can succeed.
Speaker Change: Because there is a fundamental logic that robots are an industry with a highly integrated combination of hardware and software.
Speaker Change: And the hardware system inside is very complex.
Speaker Change: It is not like a car, which mostly has real structures and such.
Speaker Change: It has many mechanical structures inside.
Speaker Change: And the progress of mechanical structures in the hardware system is not supported by Moore's law.
Speaker Change: Software can double its performance or habits cost in 18 months.
Speaker Change: Hardware has to progress gradually cars have been around for over 100 years right and there is also the revolution of Smartcards. This time.
Speaker Change: So I think in the future anyway, we are firm in our own path, which is to continuously combined with scenarios based on humanity, and then add some robotic arms and some features in some scenarios with urgent needs and pain points to complete some.
Speaker Change: High quality and highly reliable actions to achieve greater expansion.
Okay. Thank you.
Speaker Change: Yeah.
Speaker Change: We have known is that the company has a large amount of net cash on its books.
Is there any plans for share buybacks or dividends in the future.
Speaker Change: Let me answer this question and the company's CFO Thomas <unk>.
Thomas Kim: Thank you for your question.
First of all Cheetah has always had an open attitude towards shareholder return. We the management also attach great importance to shareholder returns historically, we can see that we have distributed dividends twice and have also carried out more.
Thomas Kim: Total share buyback plan.
Thomas Kim: Those two dividend distributions were based on the exit of our important investment projects and obtaining cash returns, which were then given back to our shareholders.
But for the future, whether there will be dividends or other means in fact, there are many factors that we need to consider for example, as our Vice President mentioned earlier, we actually still have some investments to make in this technology, including our business is also transforming from.
Thomas Kim: See two to be the development of AI large models and robot business also requires certain investments at the same time. We also feel that currently the overall economic environment is rather uncertain.
Thomas Kim: For us maintaining a relatively sufficient cash reserve is also quite important for the development of the business.
Therefore in the future we will continue to maintain a relatively cautious financial strategy ensured that the company has sufficient flexibility and risk resistant stability when facing market fluctuation if in the future our board of directors approved the plans for share buybacks.
Thomas Kim: Dividend, we will make an announcement to the market as soon as possible. Thank you.
Speaker Change: At present, the development of large models in China is rapid from their perspective of performance and efficiency such as the accuracy of content generation Gen.
Speaker Change: <unk> speed and reasoning cost have you felt any significant differences among various models in actual operations or past the competition of large models gone beyond the model capabilities themselves and extended more towards product innovation and ecological construction.
Speaker Change: You should know that the domestic ecosystem is rather complex and it's not appropriate for me to evaluate which model product is better.
Speaker Change: For one can try them for themselves and have their own feelings IR.
Speaker Change: I, often use them back and forth.
Speaker Change: However, it's true that there are some differences in the experience well I think your question today is very good.
I think the competition of large models has exceeded the model capabilities themselves from the very beginning.
Speaker Change: Cause the ability of the model itself as we recently launched a function called this data treasure you will find that today the ability of the model itself depends on the data and today. This data because it's relatively public on the internet when it comes to teaching and.
Speaker Change: Such if you spend more effort on this high quality data and do more engineering work your model ability can be quite good just like we actually have two models. One is with 4 billion parameters, a 17 by 881 and the 'twenty one by eight mall in fab.
Speaker Change: The results on the list are also quite good because we are here in fact, the competition of large models today.
Speaker Change: If steel mainly comes from your attention to this matter and the investment in data.
Speaker Change: And I think at this stage, it's certain that today's competition must move towards product hydration and ecological construction whoever.
Speaker Change: Whoever can truly increase user satisfaction through the improvement of the product experience can almost eliminate the gap in some indicators of the underlying model.
Speaker Change: I think there's no longer a effects the matter.
Speaker Change: Look at the list today is one is on top tomorrow down one each on top and then another one comes up for a while.
Speaker Change: In fact, I think this matter may have passed it's already moved towards productivity action and ecological construction.
Speaker Change: In a recent interview I saw that when someone asked Charlotte mom what was lacking he said it was the product right for this large model. This AI competition I think the speed is very fast. So now it has shifted from the initial technical enthusiasm or.
Speaker Change: Co comparison to product comparison and ecological comparison.
Speaker Change: This is a very obvious stage, so I see that recently, some domestic ones or entrepreneurs have mentioned that next year will be the year of ecological explosion and application explosion in principle I agree with this point because the model's ability has re.
Speaker Change: Reached a relatively high level and there isn't much difference among everyone and it's not easy to improve further.
Speaker Change: Just like the data issue, we mentioned earlier and last there is some special new paradigm.
Speaker Change: Then now the effort will be made towards the product and the ecology and.
Speaker Change: And because it has reached a certain level and everyone is investing more attention will be paid to the user experience points well. So I think there will be significant progress in terms of products and ecology next year. Thank you.
Speaker Change: I'd like to ask a technical question about large model training.
Speaker Change: How is the technical capability of Chinese enterprises under the new paradigm of reinforcement learning in large model training.
Speaker Change: We understand that the new paradigm of reinforcement learning is characterized by the lack of readily available open source models and academic papers for direct reference.
Okay.
Speaker Change: Okay.
This is really quite academic.
Speaker Change: Firstly I think under the paradigm of reinforcement learning the technical capabilities of Chinese enterprises are not bad reinforcement.
Speaker Change: Reinforcement learning has existed for a long time right I think mainly because of the launch of all benign everyone discovered that reinforcement learning might also be needed in language models right. You can look at the past after Alpha go play go some.
Speaker Change: Major Chinese companies or teams also did well in claim go basically as I. Just summarize maybe there are just a few 0.1 or two point difference in some specific evaluation indicators I think this doesn't really affect the final productive.
Speaker Change: Nation, and implementation, including our speech recognition and previous visual.
Speaker Change: Okay.
Speaker Change: We can see more and more merchants in China using service robot. The most common places our restaurants and hotel I would like to ask what is the current market share of Cheetah in these two scenarios from.
Speaker Change: The overall market perspective, how much has the robot penetrated the restaurant and hotel market how much growth space is there in the future how.
Speaker Change: How much market share can cheetos robots capture in these two segments in the next two to three years.
Speaker Change: Well regarding the ranking of market share because there are relatively few reports in this.
Industry and such they are all rather sketchy of course for restaurants and hotels, we started relatively late because we initially focused our reception services I think we should be among the top few it.
Speaker Change: It should be approximately such a share situation.
Speaker Change: From the overall market perspective, the penetration of robots in restaurants and hotels is currently still in the early stage.
Speaker Change: I think in China, it might be around 5% for a hotel, which might be a bit more because there are not as many in restaurants.
Speaker Change: It is still in the early stage with future growth in foreign countries. I think it is even earlier this time during my interviews in Europe with many of our customers right. This shows that the market space in all aspects is still very very huge you can.
Speaker Change: Imagine the foreign market as the situation, we have three to five years ago today in China, some markets do not solely depend on the robot market itself right.
Speaker Change: For example, as we all know in restaurants. It is about cost reduction and efficiency improvement and there are also some situations where their purchasing power has declined this is a fact.
Speaker Change: But in the long term I am very optimistic about these two scenarios, including some extended scenario.
What we do is not just delivery, although we are enhancing the voice capabilities using large models in this aspect of voice when it reaches a certain level. We believe it will greatly expand the work in restaurants and hotels.
Speaker Change: Oh I.
Speaker Change: I don't have this confidence, but can we achieve it based on the execution ability of our own team.
Speaker Change: This is my biggest ROE ization in the past year or so.
Speaker Change: Because it is essentially a <unk> system a lot of effort is needed in building the sales channel sales management and the establishment of the sales team. This is also why I often visit many customers now both overseas and domestic because I think our technical.
Speaker Change: All capabilities are definitely among the top in this industry.
Then, it's about which scenarios, we enter and watch products, we make and in terms of volume and customer satisfaction in all aspects, we can do well.
Speaker Change: In fact, the real difficulty here is not the technology and products for US more is the establishment of the sales channel and the entire sales network.
Speaker Change: Regarding this wave of Chinese enterprises going global I talk to a friend yesterday I think the biggest difference from the last wave of Cheetah mobile.
Speaker Change: P P going global is that we must go deep.
Speaker Change: Do some local channel construction and understand the local market and be able to formulate the corresponding strategy. If we can do this I believe our mulch will be much deeper than the last wave of Cheetah mobile's ways of using online advertising and ranking.
Speaker Change: To promote but still the effort and difficulty required for this will also be relatively large.
Speaker Change: This is our view, but I still have confidence in achieving the top position in the market within three years.
Speaker Change: Okay.
Speaker Change: Ladies and gentlemen.
Speaker Change: At this time, we will conclude our call.
Speaker Change: Gotcha.
Speaker Change: And at this time the conference has now concluded we do thank you for attending today's presentation.
Altus Kenmore: Altus Kenmore.
Altus Kenmore: [music].