In the past two months, artificial intelligence technology has been successfully applied to life scenes. Some people say that ideals and feelings have ripened the AI ​​industry, so how far is artificial intelligence to make money? The government's call, the market drive, is constantly accelerating the process of artificial intelligence. How long does it take for humans to actually consume AI?
First, iPhoneX's 3D face recognition unlocked the technology powder; then AlphaGo Zero (translated as Alpha Dog Zero), without any help of human chess, relying on their own hands and hands to learn from each other Invincible in the world, shocked the ordinary people; finally, the world's first intelligent rail transit train started in Zhuzhou, Hunan, achieving unmanned driving and inspiring the old patriots and little pinks.
Not only that, but the AI ​​industry, which has already been in full swing in China, has also injected a strong shot. The large and small practitioners have once again strengthened their inner choices--the artificial intelligence has a promising future. This industry is indeed worth investing.
As the "Holy Grail" of the computer science community, artificial intelligence was predicted as the core technical representative of the fourth industrial revolution in the 2016 World Economic Forum report, which led to the staking of Internet giants and capitals at home and abroad. A new enthusiasm for entrepreneurship. In the past two years, Chinese capital and entrepreneurs have been passionate about the AI ​​industry, and have no regrets, creating a thriving domestic artificial intelligence entrepreneurship.
Everything looks good, but there is a real problem in front of all entrepreneurs: ideals and feelings have already ripened the AI ​​industry, so how far is artificial intelligence to make money?
First, can AI finally land to create value?If you understand the history of artificial intelligence, you can understand why today's entrepreneurs and capital are so crazy about AI.
Like VR, artificial intelligence is not a new concept. It has undergone decades of development, alternating between several bonus periods and ascetic periods.
1,60s: the first dividend period + the ascetic period
The first dividend period of artificial intelligence appeared in the 1960s. Scientists at that time were confident and crazy. "In 20 years, the machine will be able to do all the work that people can do" became the mainstream voice of the scientific community at that time.
In the 1970s, because the predictions of artificial intelligence could not be fulfilled, research funding was interrupted and entered a trough, and then entered the ascetic period.
2, 90s: The second dividend period + the ascetic period
The second dividend period of artificial intelligence appeared in the 1990s. The typical sign is that IBM's "dark blue" defeated chess world champion Kasparov, and the impact is no less than today's AlphaGo's Go war. At the beginning of the 20th century, he entered the ascetic period again.
3, the moment: the new dividend period
At the moment, it is another bonus period of artificial intelligence. On the one hand, artificial core algorithms such as image recognition, deep learning, and speech synthesis are maturing and have begun to try a large number of commercial applications. On the other hand, the research of artificial intelligence has gone out of the laboratory, and technology companies have become the mainstay of artificial intelligence. Promoter.
Almost every past ascetic period, through the continuous trial and error of new research methods and logic, artificial intelligence researchers will bring surprises, such as cybernetics and early neural networks, new logic and modal logic, Prolog language. And expert systems, Nouvelle AI and embedded reasoning.
Different from the past, this bonus period is characterized by “commercializationâ€. Whether it is an industry giant or an entrepreneur, it has two roles, namely researchers and practitioners of artificial intelligence technology, which means artificial The intelligent ivory tower is getting farther and farther and is becoming more practical science and technology. At the same time, the active participation of various capitals has also promoted the artificial intelligence to enter people's lives and serve humanity.
Humans are finally going to really consume AI. Can this dividend period really create value for human daily life and bring dividends that can be realized for businessmen and capital?
Second, how many bubbles are there in the domestic AI industry?The government's call, the gathering of capital and the pursuit of the media are constantly urging the industrialization of artificial intelligence in China. Just like the wave of entrepreneurship in the past, among the AI ​​crowds, people who have already joined or are about to join AI are excited, confused, and blindly moving forward in the exploration.
Due to the surge in scholars in the field of AI, the subject education system around artificial intelligence is being formed, and formal colleges have opened artificial intelligence majors. The crash courses of various non-classical artificial intelligence vocational trainings are also “suddenly like a spring breeze, Qian Shuwan "Pear Blossoms" is reminiscent of the "Internet Thinking" training class that bloomed a few years ago: it is also between the shackles, I don’t know where to come from a lot of Internet experts, and there are many famous entrepreneurs sitting under their classrooms. The students remembered the notes with meticulous attention.
The enthusiasm is similar, the madness is similar, but the difference is that compared with the other training courses, most of the AI ​​training students are science and engineering students, and most of them have some AI related knowledge base and work experience, such as doing Over the farm. In the 80-day crash course with a tuition fee of nearly 20,000, the registration situation is extremely hot. Many people must wait patiently for the schedule to be able to win after they have paid the money, and the faculty and students are also working very hard. On the weekend morning, often At 9 o'clock, the lectures outside the training class can already hear the lectures of the lecturers in the classroom.
Many training courses have stated that they not only talk about technology, but also recommend employment for free. The relevant cooperative enterprises will come to recruit, usually there are more than a dozen companies that have booked dozens of students before a class has graduated, and students are in short supply. It goes without saying that the popularity of artificial intelligence education is due to the fact that a large number of entrepreneurs are eager for this field.
The development of any emerging industry will inevitably go through the process of being swarming, overheating, and slowly calming, focusing, and finally maturing. The occurrence of the bubble means that the industry has a high degree of attention and is a rich sea area. It also shows that the industry has a large space for falsification, and it is necessary to leave those true heroes through the big waves.
Wei Zhe, the former CEO of Alibaba, once said: "The bubble of artificial intelligence is huge, the media touted, the market is overheated, and the welcoming situation of "VC is too late, PE is too early." Many companies in the market claim to be "artificial" Smart" companies, but there are very few companies that actually have artificial intelligence technology. It is estimated that 90% of artificial intelligence companies are "pseudo-artificial intelligence."
Indeed, O2O, VR, sharing the economy... No matter which turmoil in the past, there will be some fake entrepreneurs who hang on to sell dog meat and swindle investment, artificial intelligence is no exception. Many startup companies advertise "artificial intelligence" in their business plans. It seems that they don't want to label themselves with such labels. It's embarrassing to see investors, just like a few years ago, if you don't mention O2O in BP, you will feel embarrassed. Some people just do some data, it is called big data, and then it is called big data modeling, big data risk control and the like, in fact, those data is likely to be outdated data from the Internet. Other companies have artificially implanted elements of artificial intelligence into the product in order to increase the company's brand attention, increase product sales, and even increase the price of the stock.
But these drowning fishermen have existed in every vent, and will continue to exist in the future. If these pseudo-entrepreneurs and false advertisers are regarded as bubbles in the field of artificial intelligence, the existence of these bubbles will not affect the sustainable development of AI technology, nor will it affect the determination of true entrepreneurs to apply AI to life scenes. However, it can be really affected, but it is the judgment of investors and investors. For example, Stormwind's stock is a typical case of stock speculation.
Therefore, the bubble is actually a financial concept. The bursting of the bubble is a change in the business cycle and a normal phenomenon of the human financial social mechanism. The emergence of the bubble cannot obscure the fact that the machine intelligence is developing rapidly.
And artificial intelligence is not only a vent, but also a historical trend.
Compared with VR, which also uses technology cards, the base of artificial intelligence is much thicker. The big companies that want to do artificial intelligence are far more than VR. The profitability of artificial intelligence in the B-end has been confirmed by the market; the industry prospects and money scenes involved in artificial intelligence are also beyond the reach of VR. Taking credit business as an example, JD Finance's machine automation lending is a successful case: there is no manual credit review on the whole platform, and the variable cost per order is almost equal to zero. Jingdong Financial Chen Shengqiang said: "The supply chain financial product 'Jingbaobei' can achieve 3 minutes of lending, and the trading system has the ability to process hundreds of thousands of transactions in one second, which is unimaginable in the past financial services." Since the variable cost per order is almost zero, companies have the ability and willingness to engage in inclusive finance.
For AI entrepreneurs and practitioners, instead of focusing on financial phenomena, it is better to think about how to understand the laws of artificial intelligence development, and then use them to seize the opportunities.
In today's annoyed field of artificial intelligence, machine learning, deep learning, data mining, data science, big data... these popular words are unclear, and the technology they involve is more Dazzling. For scholars who want to enter the industry, learning R or Python? Spark is so hot, Hadoop still have to learn? Popular deep learning frameworks such as Caffe, Tensorflow, PyTorch, Keras, MXNet, which one? Even for students born in the artificial intelligence class, there is no such confusion.
So which segments of artificial intelligence can be better commercialized for AI entrepreneurs who are hungry for talent and technology? Which ones can be better transformed into consumable products?
I am afraid that current entrepreneurs are more stunned and more entangled than practitioners. Today's headline founder Zhang Yiming once said: "How high is the level of talent, how high is our salary." But which talents meet the demand? Ni Hao, vice president of medical research at Yitu Technology, said, “As we often say on the Internet, we don’t know where the pigs are and where they are wool, but as long as artificial intelligence can help, it must be able to receive money... ... At present, our most determined standard is that artificial intelligence should bring value." For what kind of talent can bring value, the entrepreneur's own ignorance leads to the company often catching what kind of talents are thrown into practice to test As a result, the cost of trial and error is too great, and it has become a major obstacle to entrepreneurship.
In fact, as long as you understand the relationship between artificial intelligence and big data, the choice of startups for talents and target markets will be much clearer.
Any development of intelligence requires a learning process. The recent advancement of artificial intelligence in the past is not the result of the rapid development of big data over the years. Thanks to the development of various sensors and data acquisition technologies, we have begun to have massive amounts of data that were unimaginable in the past, and at the same time, we have begun to have in-depth and detailed data in a certain field. These are all prerequisites for training "smart" in a certain field.
In other words, the main reason for this wave of AI dividend is: On the one hand, the big data accumulated in the Internet era has reached a critical point, which can support the calculation of better learning results; on the other hand, the computing power of the computer is excessive. Used for calculations; in addition, Google and cloud computing companies and the industry have accumulated a lot of experience in distributed cluster computing. It is these synergies that have led to the development of artificial intelligence.
If we regard artificial intelligence as a baby to be fed, then a professional, massive, and in-depth data in a field is the milk powder that feeds this genius baby. The amount of milk powder determines whether the baby can grow up, and the quality of the milk powder determines the level of mental development of the baby. Therefore, the development of any machine intelligence can not be formed immediately and immediately. Even if the Apha dog does not rely on the human game, it depends on the programmer to input the basic rules to learn from it, and then simulate the massive amount of it. The game is summarized into data and rules are drawn from it. In fact, the method of defeating Alpha Dogs and Zero is very simple, that is, changing the rules but not updating the data for them. Therefore, those who can truly master the algorithm and apply it to the real business environment are the most demanded talents in the AI ​​market.
In the same way, we can also conclude that only the market segments that meet the following conditions are the industry sectors that are worth investing in Ai and making it easier to realize business realization:
1, the industry sector with a large amount of high-quality data available
The quality of the data refers to multi-dimensional, multi-form and accurate and effective. Not a bunch of data will certainly produce value, which is why Linkedin is acquired by sky-high prices.
2. Industry sectors with a large number of cross-border bilateral talents
While it seems that major companies provide both hardware (cloud computing) and software (various interfaces), in reality only cross-border bilaterals who are proficient in machine learning and understand business and users can create value in practice.
The most common situation now is that capital flows to a lot of people who are engaged in machine learning. They learn more from deep learning, but they often develop a bunch of flashy things. Most of them can only be called as an api, but they can't do one. Complete solution. Although the product looks amazing, few people are willing to pay for it because the product is too far away from the user.
People who understand the business and understand the user, although they understand the customer's needs, often do not reach the technical threshold of artificial intelligence. First of all, machine learning is not an easy subject. The understanding of machine intelligence takes time to precipitate. Secondly, artificial intelligence is not just an algorithm. To solve a lot of barriers to understanding and become a solution needs to integrate a lot of knowledge.
3, have a lot of life or work application scenarios, can form the product area just needed
Specifically, the product must be applied to closed and controllable scenarios, to assist humans in completing repetitive specific tasks, and to have achievable entry points in the application.
Take the customer service sales field as an example. Customer service is an indispensable and important role in the e-commerce era. Even a Taobao store with a monthly flow of only 50,000 to 600,000 often requires a customer service team of more than 5 people. In fact, there are often a lot of repetitive work in the customer service scene that requires too much manpower, which is a waste of resources for the enterprise. At present, Ali, Jingdong, etc. have introduced artificial intelligence into the customer service system, and there have also appeared third-party intelligent customer service cloud services such as NetEase Seven Fish and Udesk. The prospects are relatively optimistic, especially in many vertical industries with data thresholds.
4, the industry sector where data costs are not very expensive
The value of artificial intelligence applications = the value of data generated by your machine intelligence - the cost of data
The value generated by machine intelligence = the cost paid to the user for the specific service (user's attention + time + money)
Cost of data = collection cost (direct acquisition or purchase) + analysis cost (cost of hiring data scientists)
Needless to say, if the cost of data is too expensive, the profit will be relatively reduced, and it will be difficult to realize.
Four, 2B or 2C? This is a problemAlthough the ideals of the government and the feelings of the technical houses have artificially burned artificial intelligence in China, artificial intelligence is still in its infancy in China, both in terms of technology and application. Regardless of investment or entrepreneurship, its technology and funding threshold. Both are higher.
At present, it is a relatively mature institution for artificial intelligence research. In addition to national research organizations dedicated to military and national strategies, only Huawei Research Institute, Baidu, Ali, Tencent and other research institutions are in China, and relatively mature artificial intelligence technologies are only in the direction. Language recognition and natural language processing techniques, face recognition technology, mining and calculation of underlying big data, and machine learning and neural network technology.
From the actual situation of entrepreneurship and investment, enterprises must be the first to survive in the 2B field, such as robots for airlines and robots for logistics.
Taking the smart investment product as an example, the advantage for the B-end is obvious: it can fully rely on the kernel algorithm technology, and can fully satisfy the user's needs in the product experience by virtue of the abundant customer service experience of the B-end. For the C-end similar products, not only can not escape the fate of the cold start of cutting-edge technology products, to gather popularity from scratch, the logic of product experience will be relatively poor and weak. In contrast, the B-end users can precisely avoid weaknesses in these two aspects: the original user base of the organization, or the offline physical stores have given pure Internet-based smart investment a chance to start from the "giant shoulders" . The most important thing is that in addition to the number of "hot start", the foundation and accumulation of the B-end itself can greatly reduce the intelligent investment and ill-health caused by the mismatch between AI and the frontier technology and user concept: the cost of obtaining customers. On the basis of enough users, the marginal cost of getting customers on the one hand will be lower and lower; on the other hand, the willingness of new users to enter will be relatively higher.
SHENZHEN CHONDEKUAI TECHNOLOGY CO.LTD , https://www.szfourinone.com