At present, the developer platform has become a battleground for major manufacturers. Google Developer Platform, followed closely by Baidu’s AI Developer Platform, HKUST Xunfei Open Platform (released on an auspicious day of 1024). In the era of artificial intelligence, even the hardware manufacturer Sugon has started to build a developer platform (on October 24, Sugon released a series of artificial intelligence software and hardware platforms. It provides intelligent software platform services based on its own hardware servers).
The Aligenie developer platform has refined some areas:
First. The third-party content application side invokes access, and invokes its content through the developer platform.
second. Application function call. The application provider provides KFC to unify the third-party application service interface on the platform. It is convenient for developers to call. For example, if you develop a product for takeaway service, you can easily call the KFC interface to place an order.
Third, call on the device side. Midea air conditioners use Alibaba Cloud to adjust hardware devices (home appliances).
The following is the Q&A of Qianxue 2017 Yunqi Conference, head of Alibaba AI Labs (Ali AI Labs):
Question: Can you give a simple example of the Aligenie developer platform?
Qianxue: For example, the ability of " Tmall Genie " to order takeout has now been connected with KFC, which means that users lying on the sofa watching TV and getting hungry can call " Tmall Genie " to order fried chicken.
For example, in the spring of 2018, a somatosensory Thomas the train will be officially launched. In the past, it would be good to install a light that can glow and change color on a children's toy, or it can beep twice. But this little train has the ability to communicate. If you say to it, "Little train, come here", it will really come.
For example, Tmall Genie has joined the Tmall Genie store "Weijun Supermarket", which was popular a while ago. It can tell the owner that it has just received 99 yuan; Where; the store manager can ask "Tmall Genie" how much milk is still in stock. This means that the AliGenie system can be seamlessly connected and connected with the store's POS system, net sales and inventory system, and customer service system.
Question: What are the differences between Ali Artificial Intelligence Lab and Dharma Academy?
Answer: The main technical research directions of Ali Artificial Intelligence Laboratory include: speech, language, and machine vision. In terms of specific product selection, it pays more attention to human-computer interaction. The Dharma Academy focuses on a relatively broader range of technologies, including data intelligence, financial technology, natural human-computer interaction, and the Internet of Intelligence. Dharma Institute is an independent research institution outside the Ali system, and the Ali Artificial Intelligence Laboratory is a laboratory within the Ali system.
Ask: The scale of AI experiments
Qianxue: First of all, to answer the scale of our artificial intelligence experiments, Alibaba will make greater moves in the entire technology research direction. Our team is not too big at present, about 200 people, but we are still in the process of very rapid growth. We have also recently joined by several scientists.
Because artificial intelligence technology itself is naturally related to algorithms and application software. Hardware is some of the points we are trying to innovate today. Because hardware itself is a good carrier, it can make it easier for everyone to understand the level of our technology today. But not all technologies must rely on hardware. For example, today’s AR platform may not need to be directly related to hardware, because mobile phones are popular enough, and we don’t need to wait for AR hardware carriers to come out, so we probably will be like this , so we have many labs, each with a different focus.
Question: Is there any obvious focus now?
Qianxue: At present, my branch has not been announced to the public. It is true, but we have a Beijing artificial intelligence laboratory center, which has been established, and there are already some people in it. So in terms of geography, we currently have two places in Hangzhou and Beijing. In fact, part of our team is based in Shenzhen. We will be very flexible in terms of organizational structure, and we will also develop in multiple regions and fields. I also hope that everyone will pay more attention to our artificial intelligence laboratory, and we will have more actions, and you will see it more and more clearly in the future.
Question: Will there be further hardware releases?
Qianxue: What we released this time is under the strategy of the Internet of Intelligence. What Mr. Ma mentioned touched my heart very much, because we talked about the Internet of Things a long time ago, and did not include artificial intelligence or highlight it. The concept of online. Maybe everyone was surprised today. We made another Tmall router. Today, the router industry has to be subverted . This industry has existed for a long time. You can connect to a network and use it on the market, but I haven’t thought about it. What kind of routing do we need in the Internet age. So based on this, our laboratory is relatively innovative, so we chose such a product. But for products in other fields, we haven't considered yet, and I won't make some functional products, such as electric fans, which may not be my core point yet.
Another one is that because Alibaba itself has accumulated a lot of technology, the products we released today are all for Alibaba’s own use, because people used to say that Ali is a commercial company, and never thought that Ali is a technology-based company. Today, we are re-using our own technologies for commercial applications, such as enterprise-level routing, and the deployment services in the entire Yunqi Conference have already used our enterprise-level experience. We just added some technologies related to machine learning at this level for home application scenarios. As I said, the network speed can be dynamically adjusted, application scenarios can be recognized, and so on. Presumably that might be the case.
Question: Just now we saw Ali Huoyan. Is there any connection with Yitao Huoyan?
Qianxue: Of course not, I just used the word Huoyan. Because first, Huoyan itself is a pretty good name, and I like it very much. Around 2011, in fact, AR was still very immature. The part we are doing today has been developed in the past three years, because you should have seen Alibaba’s AR technology at the Creation Festival. The interactive technology of Tmall's Interactive City Carnival is used. This part of our technology comes from here, and it should be something from two eras. It's just that I think this name has a certain user awareness, so I continued to quote it.
Question: AR technology is already very mature, why not expand it to glasses and so on?
Qianxue: I personally judge that AR is like this. It can be used in many scenarios. Although AR glasses have not yet been fully realized, Alibaba has a very rich customer base. For example, Midea may tell us that it is very simple. When it When a new product is made, will users use it? The application of AR to this industry can bring great changes. Another example is the toy industry, where you usually put an image there, you don’t know what will happen to the toy in the end, but you will know it with AR technology. Therefore, we think that the core demand of users is to learn more and experience better through AR. He does not need to wait for the next generation of hardware to come out before doing it.
Another reason is that there are not too many content sources or participants in the overall AR industry today, and it cannot be said that there are not too many. It is still a top-notch technology, but it is not a company like Midea or Haier. After they are used, they can also drive AR. develop. So I think it is a process that complements each other, and there is no complete sequence.
Question: You are looking at the entire group level, such as smart audio playback, there must be music from Xiami, and video content from Youku and Tudou.
Qianxue: Obviously, it is possible. Whether it is shopping on Taobao or Xiami’s music services in Tmall Genie, we are seamlessly connected. Similarly, if it is at the level of AR, if it involves things such as Youku, it is actually quite logical that this is the case. There must be a business model in it. Youku’s content display originally has its own industry’s gameplay, and it must be related to it.
Question: The two products you released today are products with ready-made solutions. As a technology, Alibaba will come out first when it is displayed on some dressing mirrors or refrigerators we see?
Qianxue: When we were doing Tmall Genie, it was not as simple as everyone imagined, because the industry is very immature, and routers are the same. You can find routers with completely seamless autonomous networks on the market today, but now there should be no routers. You may I have only used a simple hotspot, but the effect of the hotspot is also very poor. These two are not similar at all. The hotspot only receives the signal and then amplifies it. However, our autonomous network is a network between routers. In this regard, we still lead the entire trend. Like Tmall Genie, we also feel that there is still a long way to go, and there is still a lot of room for improvement. So in this regard, we will pursue to be more intelligent.
Question: We know that there are some goods in the Taobao domain. Alibaba Cloud must have an algorithm and a basis for doing this. If you make this product today, if Alibaba can’t do it, who do you think can’t do it? Or why there is no way for Alibaba to do this? Like a dressing mirror.
Qian Xue: Because the time is quite short, Ali AI Labs has only been established for one year, and Tmall Genie has only been on the market for two months. Today, two more products are released. For me It's not easy. But I am very grateful for your expectations and confidence in us. We also hope that more artificial intelligence technologies can be applied to new hardware, so keep paying attention to us and you will not be disappointed.
Question: I have two questions about Tmall Genie. The first one is that we use the Tmall Genie system as an interactive system this time, and it is not fully open source, such as reaching Android. In addition, cooperation has been adopted in terms of sales, such as the way of entering hotels. So for Tmall Genie, what are the ideas or benefits for the development of smart audio?
Qianxue: First of all, I still hope that Tmall Genie can become a real personal assistant, so we will continue to maintain this line and this strategy, and we will continue to move forward in a relatively stable manner. Customers in industries such as hotels are really eager to have the ability to use artificial intelligence to update their current equipment and experience. They are very proactive in finding us because our platform is open enough and scalable. ability, so this is the case
After the release of Tmall Genie, too many customers came to me, hoping to cooperate with Tmall Genie in various ways, including China Southern Airlines this afternoon, who came to us immediately on the first day after the release. We talked about several scenarios. They hope to put Tmall Genie in the VIP room, and at the same time have a device that scans the registration card, which can locate and notify. why? He said it's very simple, the VIP room is quite big, and there are big bosses in it, people who often travel on business. Air China's flight delays are very serious. At present, they are looking for someone to shout for a while. After finding us, they found that our part is completely free to define upstream and downstream, and there are service interfaces to intervene in its services. On this point, it shows that our platform is doing well, and they can customize it. He thinks of it as his elf, to be more precise, it is called the Southern Airlines elf. I have it ready-made today, and I can even remove the LOGO. I think it accelerates the ability to innovate. China Southern Airlines taught me a deep lesson in this point. In the original concept, the speed of traditional industries was considered to be very slow, but in fact, there are many people with ideas in traditional industries, and they are also eager to speed up. We just shortened this time. To put customers first.
Question: You said just now that it was the first year. Yesterday, MISA was also opened and open sourced. What does this mean to you?
Qianxue: From my point of view, all this has just begun. We also welcome many people and fields to join in, because the market size of this matter is large enough and there are almost enough, so I think it is very good.
Question: For questions about developers, what are the measures that distinguish it from other platforms?
Qianxue: Take the smart home as an example. Almost all the brands that can be thought of in the market are currently our developers. They will take Tmall Genie back, because they generally have their own private cloud. Part of it will be implemented directly on our platform. So you can think about this part, if there are 1,000 brands in China, all of these 1,000 brands are doing or have already done it, in the process. Another one, like the middlemen in LT, originally like Ali Smart, Bolian, they are also docking, and Tuya Technology and so on. So I don't have statistics on these developers, but there are still quite a few. In addition to this, we have some individual developers, I don’t know how to call them, maybe his team is very small, I guess there are only a dozen people. There is a coffee shop, they have made a lot of tongue twisters and a lot of content-related applications. There is another one. There was a typhoon some time ago. I remember hearing our technicians say: There is a developer who wants to develop an APP specifically for typhoon conditions. So there should be quite a lot of people who are very concerned about our developers. How to review this part, this process is under construction.
Question: Your research direction and standards may be different from those of others, including what the mechanism of your products in the laboratory looks like, and whether the assessment is different from others.
Qianxue: First of all, our laboratory is indeed equipped with hardware and software, as well as product and market capabilities. This laboratory is indeed quite special. But the direction we focus on is more in line with the direction of Bodhidharma Academy. First of all, we will focus on the three words of today's intelligent network. This is our biggest direction. Basically, this direction is big enough. Alibaba's artificial intelligence laboratory will focus more on human-computer interaction in the selection of specific products, because human-computer interaction is also a very deep homework. For example, a simple touch screen can completely subvert mobile phones, so there are too many things to do up. So we mainly focus on intelligent networking and human-computer interaction, which are our two core directions.
At the assessment level, I still encourage innovation. I hope that what you do is different. This is my focus, and it may be different on this point.
Question: Will there be KPIs?
Qianxue: We scientists will definitely not give KPI to us. The point of view is that if your product is good enough, it will have its own traffic, just like Tmall Genie has its own traffic. Many businesses seek us to cooperate, so if you fail It might be wrong to force. I still pay attention to the fact that the kernel is good enough, and it is useless if the kernel is not good enough.
Question: Are IDST and AILAB completely non-competitive?
Qianxue: In our laboratory, taking AR as an example, we must also look for application scenarios, and finally turn it into products and services. So at this point, at least in terms of domain and model, they are still not the same. From your point of view, my technology must have a case that can eventually be turned into a product.
Question: The second question is about routers, how do you sell them? Will the sold ones be packaged with the same concept as Tmall Genie, or what?
Qianxue: I can talk about it from two aspects. First, our routers are really different from those on the market. They can self-organize networks and support Bluetooth, etc., which cannot be found on the market. Second, our quality itself is very good, which is the stability of our own routers, so I am very confident that no company will go this way. In addition, during the process of Tmall Genie, we observed that a considerable number of users are renovating and renovating their smart homes. We think this type of users is relatively related in terms of possibility. They will choose hardware products, TV sets, etc. , are such user groups. These user groups themselves will definitely not get involved in upgrading their own routers, and I really think that it is time for routers to be upgraded. Because when I was doing Ali Smart two days ago, I paid more attention to the router industry, because it is a very important link in the Internet of Things, but the routers on the market just need to be connected.
Question: I am more concerned about the segmented scenarios of Tmall Genie in the retail industry. Can you share with us more, such as what kind of application Tmall Xiaodian is.
Qianxue: The Tmall store itself is a new retail model created by Alibaba. You should be able to see information about the Tmall store online. It should provide a brand new system for the store, and there are A large amount of data guides the value. Because Tmall Genie itself is a human-computer interaction product, it only helps to improve the communication experience between store managers and customers. This is the first point. The second point is to help the store manager to manage and manage other store personnel, because in the past, a store manager had at least several employees, one for customer service, one for the distribution floor, and one for the warehouse. These can be solved through systemization and Tmall Genie. The Tmall store we built is enough for two people, this is one. In addition, the Tmall store itself has its own community and people, which is equivalent to living in this area. There is a Tmall store in this area, and this group of people is relatively fixed. We do not rule out the possibility that the things in the user's home need to be updated, and the connection between the store manager and the user can be strengthened through human-computer interaction. From the perspective of the ability to connect people, there is a lot of room for imagination. The last one is definitely right for me. Tmall Genie is in the small shop. We often greet it, and it is also very cute. Ask it where the soy sauce is.
Question: Because the AR platform is open today, our previous technology is also given to developers or service providers. Because AR including Apple and Google is also open to developers this year, everyone knows the benefits, and we all understand the imagination of the future in these things displayed. What we are curious about now is, for example, what is the difference between Tmall Labs and their thinking when they want to do this thing? What will our relationship with developers be like?
Qianxue: We are not the same as Apple. Apple has released some basic capabilities of AR, and it can pass AR parameters to you. What we do today does not require such a deep understanding and technology of AR. You can directly provide your materials to us and it can be completed. My case today is actually not very clear. The Tmall Genie is here. If you are a customer, you just need to bring the Tmall Genie. You can add 3D models to it yourself, or add anything else. You submit it yourself in the content library of my platform. Such simple two steps have already realized AR, but Apple can't do it at all. It is equivalent to developing all the capabilities of my platform from beginning to end to complete it. So at this point, we will be closer to our first-line customers. Because Apple is on the mobile phone, it can only stabilize the gyroscope of the mobile phone and so on. It's the same with Android. I think it's more difficult to do well with so many models. So we should have a clearer layering, they will be more on the hardware side, and we will be more on the customer side.
Question: For example, we have seen the first two types of display, such as displaying them on a card. The third is that I put some products and then take pictures, which is equivalent to adding a hyperlink to objects in the real world. From the developer's point of view, whether it is a merchant or a developer, in this regard, if they want to connect, they need to connect their products with the things in the store on Taobao?
Qianxue: In fact, our third case is very difficult. There is no such thing on the market. It is called 3D object recognition. Traditionally, what we see is image search. Our image is 360 degrees, no matter how you shake it, you will know what it is. Second, we have a content platform, which means that after you identify it, you can superimpose any content you want. What kind of material library do you have on the content platform, you can add links and food, it is the work content of a secondary media. This is very convenient.
Question: Is our voice platform similar to the AR platform?
Qianxue: Voice may not be the same as AR, because AR is inherently dependent on mobile phones, and currently the largest carrier is mobile phones. The voice is built by ourselves from the beginning to the end, which means that our voice is both end-based and customer-based. There are many AR models, thousands of them. At present, mobile phone manufacturers make more by themselves. For example, if you are a mobile phone manufacturer, they want to use AR capabilities to provide APP application providers on it, and they have to change their gyroscope, depth camera, etc., and they can provide better. So there is still a difference.
Question: Does our laboratory focus on Tmall Genie now?
Qianxue: In fact, the reason why so many products are released this time is because too many people think that we only do Tmall Genie. I think we still have more dreams, and we are called an artificial intelligence laboratory. So this time I also take the opportunity of Yunqi Conference to tell you that we are a laboratory, so we still want to do a lot of things.
Question: Are there any new developments in machine vision?
Qianxue: At present, the biggest progress is AR. We are also wondering whether AR technology can have more integration points with Tmall Genie or other products. Because what Academician Zhang said in the afternoon was very good. He talked about multi-modality. I think multi-modality is a way of development, and it must be the integration of all aspects, including today's acceleration chip.
Question: Since Dharma Academy has to support itself, what is our profit target and profit plan?
Qianxue: I think it’s too early now. For the artificial intelligence industry and the AR industry, the business model is still too early. I think everyone has a lot of opportunities, and we can still move us. I think artificial intelligence technology scientists have called it, and the entire human world has called it at least three times, and it has been hot three times, and none of them was completely successful in the end. This is what attracts me the most. I really want everyone to be like, wow, this is awesome, this is my focus now, and the business model is natural with this.
Question: What kind of organization rack purchase? Is there a dedicated product selection team, or is there now a team for audio and a team for routers? How is the division of labor?
Qianxue: According to the direction of the big technology field, the natural understanding of speech and machine vision have not yet completely separated the technology from the product, because the current AR technology is equal to the product. But in the development, there will be many points for the application of voice technology in the future, and it may be cut at that time. So now the organization is mainly based on the technical latitude.
Question: Many merchants are looking for you now. Now you have to deal with customers like Tmall Xiaodian. How will your team do it?
Qianxue: The laboratory is developing rapidly, and the only constant in Alibaba is change, so our organizational structure and layout are also changing every few days. The mentality of our team members is relatively good, and everyone is still developing vigorously, so it’s okay. We are also cooperating with Alibaba Cloud itself. Alibaba Cloud also has many customers like hotels, and we don’t need to completely repeat the construction. It means that I have done a good job in my product experience, and they will make applications on my platform. The way I am now encouraged.
Question: Alibaba used to have a dedicated team to develop technology, and there was also a team dedicated to using this technology, so is your laboratory now the department responsible for the application of this technology, or the department responsible for research and development?
Qianxue: Not necessarily, depending on the product form, for example, AR is completely made by ourselves, because it is too cutting-edge, I don’t think it is certain.
Question: Do you have contact with them?
Qianxue: We should be a sweeping monk in the Bodhidharma Academy
Question: This product will have a product line, or like MINI or MAX?
Qianxue: We will make adjustments according to the situation of market users, because the pace is too fast, and we can’t figure it out by thinking about it now. One good thing is that we used to do Tmall Genie completely with our brains, but now we have real users, and users will guide us how to go forward. Before the product was launched in our laboratory, we YYed a lot of products. Tmall Genie has a lot of fields. When it is actually launched, everyone will be more focused and follow the ideas and ideas of today's users. Like takeaways, there are many single young people who really want to order takeaways by voice. So in the future, when we have customers for our product development, we will follow this flow more naturally. We don’t rule out what you said. You also provide as a customer. What do you think? Let’s build together.
Question: Do we have a plan for the second generation of Tmall Genie?
Qianxue: Tmall Genie has been officially on sale for two months, and it is a little faster to release new products in two months. But the hardware still needs to be updated, which is definitely in the plan.
Question: I would like to know, for example, whether we are in a traditional industry or an existing organization, such as a hotel, do you have any case for the running-in process of similar cooperation? For example, how do we solve some difficulties?
Qianxue: In fact, it is indeed a very complicated case. Like the hotel we settled in, we are in Xixi Paradise. Today, President Lu of Xixi Paradise laughed at himself on stage, saying that they have a lot of service personnel and security guards. coordination. The advantage is that because our products are new enough, they have never seen a scene like ours on the market, so it is easy to unify the opinions of people with different structures in this way. So I think it is because the technology and application scenarios are good enough to impress them.
The other one is still difficult. If you want to change the curtains and lights in the hotel, it also involves another supplier in the hotel. This link is really long enough. So I am proud of my team, the hotels that can be moved into are released in such a short period of time, at a god-like speed. But every industry is responsible. The same is true for Tmall stores, which also involve store decoration. There are really many things, and it also allows us to better understand traditional industries. Question: Can Tmall Genie go deep into shelf management?
Qianxue: Tmall Genie is more human-computer interaction, shelf management Tmall store has a shelf management company, I just tell the store owner how the shelves are. But it is very difficult for me to know what he said in terms of natural language understanding, because no one has ever told the machine where my things are, and I have to understand these things, and I have to understand its left, right, and top, etc. Stuff is hard. So we still have to do a lot of things like hotel definition, small shop definition, etc., to build a knowledge map, understand what he said, including the correction of speech recognition.
Question: First of all, return to a question, why did you release these three products? What is the particular connection between these three products? Is it just three of your many products, or is there a relationship behind it?
Qianxue: In fact, there is a relationship. The first relationship layer is the Internet of Things, because I was also thinking yesterday that AR and the Internet of Things are related. these three. Second, as I just said, these technologies can be crossed again. If you think about it, if Tmall Genie has AR technology, you will think it has something to do with it. So at this point, it's just because of this time today. I can't directly bring out my many, many ideas in the future, and I have to leave you with a sense of mystery.
Question: Or are these three products relatively mature?
Qianxue: It is certain to post again when you are mature. I think it’s because I believe in multimodality, and I believe that the combination of voice and machine vision has a very explosive power, but I can’t directly reach that point today, so I will first give you the first step I made. experience it.
Question: In the past two days, two platforms and two capabilities have been continuously opened. First of all, the Tmall Genie you made yourself is there, but which scenarios or main cases do you expect to apply it to?
Qianxue: This problem is not really planned by me. I think it still depends on our customers, but my own judgment, for example, smart home is a direction, because now smart home customers often use Tmall Genie to do something. Behind it may be anything related to housing, such as real estate. The other is the feeling of a robot, such as being able to communicate, and what else can you do after the communication is over. So these two are what I have in mind.
Question: Another reporter asked just now. You may be relatively familiar with the smart home, because many manufacturers are in contact with you. Are there any other aspects?
Qianxue: There are also other aspects, such as today's Thomas the train, which is a completely different brand new toy market, and AR can also be used in the toy market. If you really want to ask me about the relevance of these technologies, the customer behind them is the same. For example, today I am the customer who made Thomas the train. It was the first customer who wanted voice technology, because it wanted to make its toys more It's more fun, it needs content, so join us. Second, during the sales process, if you put a small train there, you don’t know the difference from other products. At this time, you put an AR, scan the back to see how to play and how the machine deforms, and it will come out. But on a traditional shelf, it is impossible to prepare a screen for this product. In addition, if you want to chat, you need to be connected to the Internet and need a router. So my customers are the same. They all have needs for voice technology, networking technology, and smart technology, so it is still smart networking, which will not change.
Question: Did Thomas the Tank Engine come here on his own initiative?
Asayuki: Yes. Including our smart home, there are several pain points for smart home customers. The first is whether the distribution network can be configured well. The distribution network is a big problem for smart homes. For example, if you take a refrigerator or something, the network distribution rate is often very low. The second point is that in the process of offline sales channels, we hope to allow users to show more, whether it is the appropriate size or how to use this thing when it is opened. Third, all smart home manufacturers hope that voice has something to do with smart home.
Question: Is Teacher Ma a user of Tmall Genie?
Qianxue: Of course, one day he suddenly came to our No. 7 building to wander around. The first time he saw this thing, he was shocked at the time. He ordered a Suzhou Tanping and a Jacky Cheung's songs, he also likes to ask some history.
