Artificial intelligence has become a hot topic today and will be one of the main technologies of the future. So now many people are aiming at this promising industry, and they have raised the banner of entrepreneurship. Then, artificial intelligence is a huge industry. After all, there are still huge problems in technology and experience. Therefore, artificial intelligence wants to start a business. There are six key points that must be known. To this end, this article summarizes the six core issues of artificial intelligence entrepreneurship. First of all, if everyone chooses to start a business today, I suggest that you should pay more attention to artificial intelligence, not the Internet. Why do you say this? 1. The traffic dividend on the Internet has disappeared; In terms of PCs, global PC shipments have fallen for five consecutive years. Do you know who is the last PC Internet unicorn in the country? It’s known, it’s probably launched in early 2011. After so many years, there’s no unicorn with PC Internet. To make an analogy, we know that the penetration and competition level of mobile Internet in 2015 is similar to that of PC Internet in 2011. By analogy, it is difficult to get a unicorn after doing mobile APP in 2015. After all, China's mobile phone shipments have been more than 500 million units for two consecutive years, and the growth has slowed down. On the other hand, wireless traffic has basically gone flat. If you sell one more, I will sell one less, which is stock competition. Today, entrepreneurs are making a pure Internet app. The first question investors ask is how you get customers. Because the traffic pattern has been fixed at this stage, the first screen is the few APPs. 2. The opportunities for Internet+ are equally limited; The main value of the Internet lies in solving information asymmetry and connectivity. So it is especially valuable for e-commerce. Taobao solved the information asymmetry with the credit system such as crown and diamond, and at the same time connected so many buyers and sellers across the country. This is the value of the Internet. But many industry information and connections are not pain points. Take medical examples, there are so many doctors in the top three hospitals in China. It is useless to connect the 1.3 billion people in the country with these doctors, because a doctor can only see so many patients a day. The Internet has not improved the efficiency of doctor visits. In traditional areas such as catering and medical care, the Internet's help is very limited. It also includes drip taxis. The Internet solves the problem of difficult taxis, but it does not solve the problem of taxi prices. In fact, after the subsidy was removed, everyone found that the drops were not cheap at all. The reason is very simple—whether it is a car or a taxi, or it needs to be opened by people. If the labor cost cannot be reduced, it will not be cheap. 3. It is artificial intelligence that can really improve social productivity and solve the imbalance between supply and demand; The increase that artificial intelligence will bring to social productivity and its impact on humans will far exceed the Internet. Still taking medical care, many primary hospitals are not high-level, and in the future, artificial intelligence can be used to assist doctors in reading medical images such as CT and X-ray. Like this year, IBM Watson's diagnosis of cutaneous melanoma has increased to 97%, far exceeding the average of 75%-84% of human experts. In the future, there will be no doubt that artificial intelligence will have huge social benefits in the fields of unmanned vehicles, robotics, medical care, finance, education or other fields. I think the next big trend and big dividend is not Internet +, but artificial intelligence +. I suggest that current entrepreneurs should pay more attention to entrepreneurial opportunities in the field of artificial intelligence. Artificial intelligence is mainly divided into three layers. At the bottom is the infrastructure, including cloud computing, chips, and frameworks such as TensorFlow. Above the base layer is the middle layer, called Enabling Technology, such as image recognition, speech recognition, semantic understanding, and machine translation. The base layer and the middle layer are the battlegrounds for the Internet giant. For example, in the chip field, Intel, NVIDIA and Qualcomm have invested heavily and the competition is extremely fierce. The same is true for cloud computing and frameworks. It is not a territory where small companies can get involved. Nowadays, BAT is also extremely important for the general technology of the middle layer. Because everyone believes that artificial intelligence is the next wave of industrial revolution. For the giants such as Tencent, Ali, and Baidu, if you want to stand up in the big waves, you must build an ecosystem of artificial intelligence (Ecosystem). The core is to rely on these Enabling Technology. What is the biggest advantage of BAT compared to startups? First, there is no shortage of data. Second, in order to build their own ecosystem, the future universal technology must be free of charge. Third, although the general technology is free, BAT has the chance to have wool on the body. This is a typical Internet play. What is the pig here? Pigs are cloud computing. For example, Baidu's ABC strategy stands for artificial intelligence (AI), big data (Big Data) and cloud computing (Cloud CompuTIng). AI, I can not make money, open to everyone, then everyone wants to enjoy my service, come buy my cloud. For start-ups, only general technologies such as image recognition, speech recognition, semantic understanding, and machine translation are expected to sell money through the SDK, and the road will become narrower and narrower in the future, especially under the pressure of BAT. So from this perspective, the risk of starting a two-tier startup is relatively large. I think that the opportunity of a startup is at the top, that is, holding the results of the next two layers to serve the vertical industry, which is what we call artificial intelligence. The artificial intelligence + that goes deep into the vertical industry can be subdivided into two types of situations: “artificial intelligence + industry†and “industry + artificial intelligenceâ€. There is a clear difference between them. The “AI+ industry†simply means that before the AI ​​technology matured, this industry and products never existed. Such as automatic driving, Amazon's Echo smart speaker, Apple's Siri voice assistant. There is no such product until the artificial intelligence technology has broken through. Because of AI, it has created a brand new industry chain. “Industry + AI†means that the industry itself has always existed, and the industrial chain is mature. It used to be completely labor-based and has relatively low efficiency. Now, after adding AI elements, the efficiency of the industry has been significantly improved. Such as security, medical and other fields. Objectively speaking, both categories have entrepreneurial opportunities. But "AI+ industry", because it is a new industry chain, the startup company and the Internet giant are actually at the same starting line. The giants have the advantage of data. So from this perspective, “industry + AI†is more friendly to startups and easier to build barriers. I believe that the future industry barriers are the biggest moat for artificial intelligence entrepreneurship. Because each industry has vertical depth, although BAT technology is better, it is not critical. Take medical + AI for example, what is the most important? A large number of accurate data that has been labeled by doctors is the most important. Without data, the talented scientists are useless. But at home, this medical data is very difficult to come up with. Therefore, BAT has no advantage in medical treatment, because they have to make these data from various hospitals and departments to be very tired. Conversely, if an entrepreneur has been working in the medical industry for many years, it may be easier to pick up the data than the big company. This requires that the partners of the founding team must have talents who understand the industry and have industry resources. This is the same as the Internet +. Once it is subdivided into specific industries, it does not mean that you have Baidu and Tencent have funds and traffic. If you invest in talents, you can do everything. There are also industry resources and contacts. The reason why I talked to you about this topic is because I went to Baidu University to talk to you in the previous paragraph. They mentioned Baidu artificial intelligence in the application of unmanned vehicles and DuerOS. At the same time, I asked me that the application of face recognition in the field of domestic security is very large. Like Hikvision, which has a market capitalization of nearly 300 billion yuan, the net profit per year is nearly 10 billion. Baidu should consider entering the field in terms of AI. I replied that you should never, because security is a typical "industry + AI" field with huge barriers. Even if Baidu technology is good, the face recognition rate is one percentage point higher than Hikvision (not necessarily, there are hundreds of AI R&D teams behind Haikang). But this does not mean that Baidu can replace Haikang. Because security is a "non-mission-criTIcal", I have identified 95 of 100 prisoners. You have identified more than 96 ones, which is not so important. In turn, what advantages does Haikang have over Baidu? First of all, Haikang is a camera. It is natural to run your own algorithm with your own hardware. Just like an Apple phone, the software and hardware experience is better. Secondly, Haikang has done so many years of security, accumulated a lot of data, face data, environmental data ... in the security field has data advantages. Finally, Haikang has done a lot of things for the public security system, such as the police service, base station information collection, view file management and other SaaS platforms, as well as the police cloud system. We can think of the IT system of the public security system, and some of them are involved in Hikvision. These things may not make money, but they have created barriers for Haikang. Because the underlying infrastructure is built by me, the front-end thing can only use mine (I can have 100 reasons, saying that competing products are not compatible with me). Moreover, Haikang has been doing this for a long time and has accumulated a large amount of customer resources, especially the resources of the government public security bureau. It takes time to develop these resources. These are the so-called industry depths. So even for BAT, if you want to enter the "industry + AI" field, you should be very cautious when choosing a vertical track. In the face of huge industry barriers, I really don't mean that my algorithm is better than you. The market is mine. Only the technical advantage is still far away. Returning to “AI+ industry†and “industry + AIâ€, the industry of the former is usually shallower, while the latter has huge industry barriers. Industry barriers are the biggest moat for startups and the key to resisting BAT. When it comes to entrepreneurship in the field of artificial intelligence, many people have a misconception that if my team does not have a big cow scientist, such as Stanford, MIT's doctoral sitting, I am embarrassed to start a business in artificial intelligence. In fact, this cognition is completely wrong. Because in the field of artificial intelligence, how important an algorithm is depends on which industry you are going to enter. According to different industries and application scenarios, I believe that the artificial intelligence entrepreneurship has the essence of mission-criTIcal and non-mission-criTIcal. For the convenience of everyone, we are referred to as “critical applications†and “non-critical applicationsâ€. "Critical applications" must pursue a number of 9 after 99.9%. If you can't do it, you can't commercialize it. For example, do you think that 99% reliable autopilot can get on the road? Certainly not, meaning one accident occurred 100 times. 99.9% did not work, and 1000 accidents occurred. Remember, the 99% and 99.9% reliability gap is not 0.9%, but the reverse is 10 times. It also includes surgical robots, which sounds 99.9% reliable, but means 1,000 medical accidents, placed in the United States, hospitals can not be bankrupted by huge claims. Therefore, the field of "critical applications" is the field of artificial intelligence that can't be made with a little mistake. It must be a technical bull, scientist or algorithm expert. At the same time, the development cycle of such projects is very long. Just like Israel's Mobileye, an ADAS (Advanced Driver Assistance System) solution, it was acquired by Intel for $15.3 billion in March. Do you know how long the company's research and development cycle is? Mobileye was founded in 1999, and it was 2007 that they launched their first product and earned the first pot of gold. Up to 8 years of research and development. This is unimaginable in Internet entrepreneurship. Including Google's unmanned vehicles began research and development in 2009, and has not been commercialized until now; Da Vinci surgical robots took 10 years from the start of research and development to the US Food and Drug Administration (FDA) certification in 2000. The common feature of “critical applications†is that projects are often expensive, have a long development cycle, are far from the money, require sustained financing, and how can the team continue to finance? At least have a very good resume and a very good background. This is a necessary prerequisite for continued financing. So everyone can see that today's unmanned entrepreneurial team is Gao Fushuai. Because it is not a rich and handsome, you can't find the day when the product is actually commercialized. Of course, if you are a "critical application" in the field of artificial intelligence, then most entrepreneurs have nothing to do. In fact, 95% of entrepreneurial ventures in the field of artificial intelligence are “none-mission-criticalâ€. Simply speaking, for these areas, the reliability of AI has to go through the basic line, and the difference is a little higher. In the simplest case, many companies now use face recognition for access control. You bring a hat today, wearing a sunglasses or a mask tomorrow, the recognition rate can't be 99%. But no problem even if it is not recognized. Because all access control with face recognition has a place for you to press the fingerprint. Even if the fingerprints are not brushed in, the problem is not big, the company does not have a front desk. This is the "non-critical application." This type of project does not pursue a lot of 99 behind 99%. In fact, most of the areas of domestic artificial intelligence and robotic entrepreneurship are “non-critical applicationsâ€. Of course, it is not that the algorithm is not important in this field. You can't recognize it every day, so you must pass the basic availability threshold. Occasionally, problems can be tolerated. “Key applications†cannot be tolerated. "Non-critical applications" do not pursue high, simple, practical, cost-effective and more important, such projects usually compete for comprehensive strength. include: Insight into the industry. Be familiar with industry pain points; Product and engineering capabilities. Light does not make sense in the laboratory; Cost Control. Products that can not only be made, but also have to be made cheaply; Supply chain capabilities. Not only can it be shipped, but also mass production; Marketing ability. The product is out, you have to sell it. There are no marketing experts in the team, and it is the key to get the best channel. Therefore, when you are in the startup team, you must think about which field the track you choose is in. The different track requirements for the team are different. “Critical applications†must be technically sturdy, and “non-critical applications†require teams to be more comprehensive and comprehensive. Many artificial intelligence entrepreneurs are now born from a technical background. The first idea of ​​starting a business is usually to be a technology provider. Technology providers can be a stepping stone to entrepreneurship. But if you only target technology providers, the road ahead will be very narrow. Why is it that the value of technology providers will become smaller and smaller in the future? There are several reasons for this: 1. First of all, the general technology must be the track of a large company, and BAT will be open for free in the future. People's big companies will provide EnablingTechnology such as face recognition, speech recognition, semantic understanding, and machine translation for free. How do you plan to make money by API call? Maybe you can make a little money now, but it's hard to be a long-term business. 2. The technical barriers that rely on algorithms will be lower and lower. In the future, with the rich maturity of the basic computing platform and the open source platform, the technical barriers will become less and less obvious, and the technical access threshold of the entire artificial intelligence will be lower and lower. Just as you want to find an IOS developer in 2008, it's hard, but now it's easy. All the evolution of technology follows this rule. In particular, with the computer majors of universities today, there are many machine learning courses, and there is no shortage of talents in the future, which will lower the barriers to entry for the entire industry. At the same time, with the maturity of ecosystems such as Google TensorFlow, many areas will have well-trained models for reference (the demo will be faster), and entrepreneurs only need to have enough data to train the parameters. So the barriers to future algorithms will be lower and lower. If the company's core competitiveness is only an algorithm, it will be very dangerous. 3. If the technology provider does not provide a total solution for the user/customer directly, it is very easy to be crushed by upstream and downstream: For technology providers and algorithm companies, if your technical barriers are not high enough, the upstream is likely to do your thing directly. Examples of this are everywhere, such as companies that provide Hikvision with face recognition algorithms. The problem is that when Haikang uses your algorithm, there are also huge R&D teams working on their own algorithms. If you are using it now, you are not ready yet. Once you are ready, you will be replaced immediately. Even in industries with a certain technical threshold, the days of technology providers are not too good. For example, Movidius, which focuses on embedded visual processing chips, has been using their chips. But since Dajiang ruled the consumer drone market, Dajiang is now naturally starting to develop its own chips. It is said that the technical barriers of the chip are not low, but as long as the industry concentration is high, the winner will choose to take all. For example, the manufacturer of mobile phones has reached a threshold and has the power to make chips. Like Apple, Samsung, Huawei and now Xiaomi, they have chosen to do their own mobile CPU. Therefore, these technology providers such as MediaTek and Qualcomm are actually very painful. This is actually a general rule of the industry chain: the monopolist in the industry chain will eat all the profits, and they are very motivated to expand upstream or downstream. Take the example of the PC industry chain, memory, hard drives, complete machines, monitors... do not make money. Who earned the money? Windows and Intel have earned most of their profits. Since there is no way out for a pure technology provider, what should I do? Hao Ge put forward the theory of “one horizontal and one verticalâ€. Technical services can be done in the early stage, but technical services cannot be done for a lifetime. “One horizontal†refers to the technical services you provide. Usually "one horizontal" can serve many industries, you must find 1, 2, you think that the most market opportunities, the most suitable for your vertical field, deep into the "full stack": turn technology into products, and then get users Selling, realizing business realization, and then more information on the business, more consolidating your technology. In a word, to be a "full stack" of technology, product, business and data, this is "one vertical." This is a healthy business model. In vertical industries, because there is no conflict of interest, you can still do technical services honestly. In this case, you can thoroughly understand a vertical industry in business, and technically you can form a more data loop through horizontal cooperation to consolidate your technology. This is the theory of "one horizontal and one vertical". So for technology startups, from "one horizontal" to "one vertical", which vertical area to choose depends on five key factors: Is the market space not big enough? Do you do a full stack of vertical fields, or do you want to be a horizontal technology provider? It depends on which market space is bigger. Finding the right vertical field, even if it only takes a little bit of market share, may be more profitable than doing "one horizontal". Take the example of Meitu, they have apps such as Meitu Xiuxiu, Meipai, Beauty Camera, etc. At the same time, they will cooperate with many mobile phone manufacturers to provide the beauty effect of camera shooting. You can understand that this is technical service. But after studying the 2016 financial report, do you know what the "one vertical" of Meitu Xiu Xiu is? It is a Mito phone. The technical services mentioned above are far from making money by making a beautiful mobile phone vertically. Mito phones accounted for 93% of the company's total revenue. Although the sales volume of Meitu mobile phones last year was about 748,000 units, it only accounted for less than 0.15% of the domestic mobile phone market's annual sales of more than 500 million units. What is the concentration of the industry? When you are a "horizontal" technology provider, the most worrying thing is that your upstream or downstream is too concentrated, or the more obvious the head effect, the more unfavorable the technology provider. To give a simple example, in the IDC era, HP, DELL and other vendors sell servers, which are sold directly to various IT companies. Everyone has been very moisturized. But it's hard to do after 2010, because cloud computing has emerged. The vendors that provide cloud computing will be able to count them out. And the head effect is extremely obvious, only the Alibaba Cloud family accounts for more than 50%. If you are a technology provider, you will find that there is no chip in negotiating with such a monopoly industry. So now I am very sad, suppose I am Aliyun, will let you list the BOM cost, I will give you a 5% or 10% profit, this business is very difficult to do. In this case, of course you have the willingness to go upstream too. But what is the problem? If the upstream concentration is high, it means that the barriers to this matter are very high. It is also very difficult for you to go up as a technology provider. If this upstream concentration is low or the customer is very scattered, it is a good thing for you. But you don't have much motivation to go upstream, because this market is very fragmented. Even if you kill it, you may only have 1% market share, and 99% of people will become your competitors. This is a paradox. Is technology a revolution or a revolution? If your technological innovation is revolutionary in this vertical field, the more you have the chance to go upstream. If it is only improved, you will earn a hard earned money in the downstream. The more subversive things, the more opportunities you have to go upstream. Because the upstream is inseparable from you, it means you have the opportunity to do his business. For a whimsical analogy, if you can provide a “standby for a week†battery, then you can consider making your own mobile phone. Your mobile phone only has one point: no charge for one week, and it is the only one in the world! This may be enough because the technology is revolutionary. On the contrary, if it is a modified technology, for example, your battery standby is only 10~20% more than before, then you still honestly sell the battery. Who is higher than the barriers? The barriers of technology providers and the barriers of upstream customers are higher, and they also decide to make a "one vertical" success or failure. For the live broadcast platform, there are now beauty features, such as giving girls a hair, which is usually a technology provided by a third party. The barriers to technology itself are not high and can be provided by many companies. Although there are some small differences in performance, you have no obvious advantage. However, the barriers to live broadcast are quite high. This has a network effect. The more users will attract more beauty anchors, because more money can be earned, and the more beauty anchors, the more users will be brought. At the same time, you are willing to spend money, you need a lot of money to buy traffic and sign up for a very NB anchor. So this thing has a high barrier. Your technical provider barriers are not high. In this case, although the technology provider can only make a hard earning, there is still no chance to go upstream. Does it match the team's genes? Being able to do technical services does not mean that you can do vertical solutions and do a full stack. Because the team does not necessarily have industry experience, this is a big problem. After Amazon's unmanned convenience store Amazon Go came out, many domestic technical teams also wanted to provide similar technology, and even wanted to make 2C convenience stores. After talking with them, I will advise them to think about it again. Your technology is no better. For the user, when he buys something, will he see someone in this convenience store still no one? No, this is not a priority option. His first consideration is - which convenience store is closer to me, and whether I want to buy something that is convenient. In this sense, this is back to the essence of retail. So if the team doesn't have the retail gene and doesn't know how to retail, don't consider opening a convenience store. At this time, many people may ask, "Isn't it possible to find an executive who knows the industry?" This is not so simple. If the CEO does not understand the nature of the industry, it is difficult to make up for it by an executive. I especially believe in genetic determinism. If any new business, BAT can find an executive who knows the industry, then the business of China's Internet is all BAT, and there is no business. BAT, one to do search, one to do e-commerce, one to do social. In fact, all three of them have tried the other party's things and never succeeded. So what everyone can do, what not to do, is highly relevant to the company's genes. The last question, to put it simply, is that it takes a certain amount of time for technology to mature. Because from the perspective of any technology popularization, almost all of them continue from the military (aerospace), to the government, to the enterprise, to the B2B2C, and then to the 2C. The same is true for artificial intelligence. At present, artificial intelligence is not very mature in the 2C market. Simply put, robots, in the personal consumer market, there are only four types of robots with large shipments: sweeping robots, drones, STEAM educational robots and smart speakers represented by Amazon ECHO. Why is there a certain difficulty in the early popularization of the 2C market? There are several reasons for this: 1. The industrial chain is not mature I make an innovative thing, the finished product has 10 parts. Every part has to do it yourself, and because of the small amount of shipments, there is no scale effect on each part, which makes each part expensive, so you must make the finished product very expensive. This is a very big problem. 2. 2C is extra money This is also a very important issue. The 2C end users are usually sensitive to the price because they pay for themselves and spend extra money. The product is very expensive and is a big threshold. 3. 2C products have high expectations Users buy such expensive things, and naturally expect more products. Everyone thinks that I am buying a robot back, I can't wait for anything to do: I can sing, dance, chat, clean, and speak English. But this is unrealistic, and the current technology maturity is far from this. Compared to the 2C end, these problems are not a problem at the 2B end. 1. 2B end is more affordable First of all, the ability of companies to withstand prices is obviously much stronger than 2C. You said that a robot 20,000, 2C consumers can not buy, but the enterprise problem is not big, the company has high cost tolerance. 2. The core purpose of 2B is to reduce costs. For example, an industrial robot, 100,000 yuan, sounds expensive. But an industrial robot replaces your two positions. These two jobs also have 100,000 yuan a year, not counting four insurances and one gold. Then the robot can work for 4 years, and the cost is only 25% of your original, or even less. Then the company is still very cheap when it comes to accounting. 3. 2B can adopt human-machine hybrid mode There are also 2B-end robot applications that are simpler. On the one hand, most of them are single tasks. The robot only needs to do one thing and it is simple to implement. In addition, many of them are working in the "human-machine hybrid" mode. That is, I used to need 10 people to work. Now I use robots to replace half of them. The simple and repetitive work is replaced by a robot, and the complicated use of the remaining five people is the "human-machine hybrid" mode. For example, there are many security robots at home and abroad, patrolling on a fixed route. You can understand it as a mobile camera, of course, the algorithm must definitely add something to identify. Fixed route patrols, this can be done to the robot. The difficulty is that in the process of patrolling, if you find that the old lady has fallen and let the robots get up, this is still not possible. With the rapid development of wireless sensing technology, 433 MHz wireless communication devices have been more widely used in portable devices, vehicle-mounted terminals, intelligent locks and other fields [1]. As an important part of wireless communication equipment, antenna is a key component that affects the overall performance of the communication system [2]. 433MHZ Rubber Antenna,433Mhz Patch antenna,433MHZ antenna with magnetic base,433MHZ Yagi Antenna Yetnorson Antenna Co., Ltd. , https://www.xhlantenna.com
Domestic and foreign scholars have been exploring the 433 MHz printed antenna with high gain and miniaturization for many years. However, there are two main trends in the design of 433 MHz printed antenna by previous scholars: one is to sacrifice size to ensure high gain, such as the structural scheme in literature [3]; The other is to sacrifice gain to ensure size miniaturization, such as the scheme in reference [4].
Taking into account the effective size and gain characteristics of the antenna is the difficulty in the design of 433 MHz miniaturized printed antenna. Based on the research experience of domestic and foreign scholars on 433 MHz printed antenna, a 433 MHz miniaturized spiral printed antenna is designed based on the 1/4 wavelength monopole antenna. The simulation results show that the antenna occupies only 20×35 mm2 and the effective gain is -4.14 dB.