Recently, a blog written by Bradford, a partner of machine learning and financial venture capital firm DCVC, has drawn great attention abroad. He believes that there will be five major changes in entrepreneurship in the field of artificial intelligence in 2017. This article may provide some insights for investors.
Bot will go bankrupt; deep learning will be commercialized; artificial intelligence will become a "clean technology" replica of venture capital; MLaaS will fall into a second stagnation; full-stack vertical artificial intelligence startups have developed.
As the enthusiasm for AI is slowing down, 2017 will be a reorganized year. The trend of pure hype will not break. Paradoxically, a small number of vertical AI startups will solve the full stack industry problem after meeting the relevant expertise, unique data and the need to use AI to deliver their core values, so 2017 is also a breakthrough victory. One year.
Bot will fall into bankruptcy
In the past year, there has been a mad pursuit of bots.
In the technical community, when it comes to robots, we usually refer to software agents, and are defined by four key concepts, so as to interact with any program (arbitrary programs), environment (reacTIon to the environment), autonomy (autonomy) Agents, goal-oriented, and persistent agents differentiate.
The industry has stolen the term "bot" that used to refer to any form of business process automation and created the new term RPA - robotic process automation.
Of course, business process automation will continue to play a role for the next few decades, and the robotic fanatic that now stands as a "bot" (a conversational interface with voice and chat capabilities) will begin cooling in 2017. The reasons are as follows:
The socialization and personalization dispute in the field of consumer Internet provides a good reference. The final winner is the personalized platform Facebook, which is also a social platform. People still like to communicate with people on most things, and I suspect that many chatbots will use the same approach as non-social media platforms to try to bet personalization without social strategies. Much of the thinking around the bot is superficial utilitarianism, lacking social intelligence to identify the human needs that people meet to communicate with each other. For this reason, most bots have a hard time retaining users, even if they are attracted to them at the beginning.
The outbreak of global communication apps, such as the rise of Slack and the success of specific social platforms such as China's Weibo, have released many misleading signals. Many people infer and bet on platforms such as the AI-driven digital personal assistant. According to the first article, these social platforms are addressing people's utilitarian and emotional needs. However, it is unclear to infer that it can be applied to pure utilitarian AI chatbot.
Compared to other, more visual solutions, conversational interfaces are not always as effective at completing tasks. The conversational interface is interesting and has been around for decades in the HCI community. In some applications, the conversational interface performs very well, but in reality, I think that the interface with high efficiency and high efficiency can be used in most applications.
Note that I have not said that AI is not good enough. More problems with most systems like siri are poor execution. We used modern technology to create a lot of interesting robot interfaces, and then there was a bigger problem in my mind: the robots didn't know that we wanted to use them.
Deep learning commercialization
Deep learning is very popular now. For those who don't know other AI terms, deep learning is part of machine learning, and machine learning is part of artificial intelligence. Deep learning is not a new thing. It is just a series of cool work that provides the best answers to many important questions, and people can benefit from it correctly.
Deep Learning Startups have replaced iOS mobile app startups five years ago. Many companies are surprised by the ability to learn deeply, especially computer vision that produces superior results and solves new problems. As a result, we saw Google, Facebook, Twitter, Uber, Microsoft and Salesforce actively adopting M&A strategies to fill vacancies.
So if deep learning is so important and highly sought after, why do I think it will be commoditized this year? The reason is the 2016 NIPS meeting and all other meetings. It is clear that deep learning is now ubiquitous and there are many graduates in this area. The situation four years ago was quite different. Today, the market has made adjustments to create more talent.
Now, I want to make a clear statement on what I said above. I think deep learning will become a bigger community in the machine learning crowd this year, but I don't say that machine learning itself will be commoditized. Machine learning talent is still hot. The benefits of the acquisition of deep learning startups we've seen in the past few years will collapse after the second wave of technology companies and external technology companies (such as Detroit's) complete the current wave of acquisitions. I expect a group of stable late arrivals to enter with stupid money this year, but we may see a wave of mergers and acquisitions slowing down.
Artificial intelligence becomes a "clean technology" replica of venture capital
Let us recall the main reasons for the recent bankruptcy of Cleantech, which applies equally to artificial intelligence.
Clean technology is not a market. It is a cross problem. Climate change and sustainable development are very serious issues and definitely deserve to be taken seriously by people as a career and profitable business. The cross problem is not a business, the business is to provide products or services, consumers want to buy. Tesla and Solar City are undoubtedly successful cases in the field of clean technology, but it is important to note that they are full-stack businesses – a car company and a solar company. So when a full-stack company with clean technology elements sells real products to the real market, this is feasible, but clean technology for purely its own purposes does not work. Because it can't meet consumer demand. Great business is the starting point to meet consumer demand. Great mission-oriented business begins with a vision defined by consumer needs and includes tasks that fully meet the needs. An organization with a social mission that does not meet the consumer's needs vision is at best a fairly effective charity. Great business puts consumer demand first, not cross-tech trends, even if it has a sense of mission.
Green energy is not a market, but energy is. Solar energy has become the market's number one and has grown rapidly due to its economic operation. When Buffett and Musk compete in this market, it is likely to show that this is a good business. Both sides see sustainability as a mission, but at the same time understand that it makes sense to think of it as a business and to prioritize consumers. The mission can only be fulfilled in a service that meets the needs of consumers and employers. It’s more ironic that there is no business with a sustainable mission that is not sustainable.
Self-respect and save the world. The clean technology is full of typical arrogant technical fanatics. In the past two years of artificial intelligence development, we have begun to see self-expanding ethical committees of artificial intelligence, and people who discuss how machines do everything, how we do it, and so on. It is precisely because we are working on a crucial thing that those who work in the artificial intelligence circle have the responsibility to lead the human development process. Pride arrogant people do not see this fact: they are deeply immersed in such an echo room, where everyone cares about technology trends rather than consumer demand and business economy. It is this harmful reality that distorts the field and brings many intelligent but self-important people into the impending doomsday of the Internet.
Clean technology and artificial intelligence are profound technical issues. Entrepreneurship and venture capital communities that are often instilled with the Internet and trivial SaaS services are increasingly having difficulty assessing investment opportunities in the deep technology arena. Driven by the pride state outlined above, after reading a blog post and listening to a few words, I plunged in. The files on Linked are updated randomly, and the era of a temporary expert is coming.
So how does this happen?
I have a theory that the economic information age has fundamentally changed the cycle of fanaticism-terror that we have experienced in human history. As a former hedge fund practitioner, I read all the famous books on financial history and market psychology. It is interesting to explore how things have progressed in different directions since the mid-1990s. I think the dramatic increase in social activity and online information expansion has created a self-heisenberging effect that pushes the business cycle before it begins. The consumer Internet is a huge example. At the beginning of the real economy, the pre-fanatic fever of the 1990s led to the 2000 crash. Two years later, in 2002, Google, a company registered in 1998, hired all the talent at the bottom of the economy and defined the true business cycle of the consumer Internet.
In the four years that Wired magazine announced the death of clean technology, solar energy has been the most environmentally friendly and cheapest energy resource. Both Mask and Buffett are keen on it. Tesla and Solar City became a full-stack cleantech empire. So I think we are on the eve of the enthusiasm for artificial intelligence startups. The vast majority I have seen is failing in the same way that artificial intelligence startups have been failing for 10 years. This is a small community of people with more than 10 years of experience in artificial intelligence startups.
This group of people at the top of the eve of the frenzy is repeating the same mistakes in clean technology. They only have artificial intelligence in their eyes, and there is no consumer demand.
The vast majority of today's artificial intelligence startups are nail hammers. This will become more apparent in the next 1-2 years. Big companies are exhausted and reduce the need for artificial intelligence talent, just as they do for mobile application developers. I guess we are starting to see the founders and venture capitalists realize that something is coming to an end. At this point, I will hear less about the decision to join an artificial intelligence startup in the past year on linkedin.
MLaaS will fall into a second stagnation
Turning machine learning into a service is an idea that we have been thinking about for nearly a decade, but this idea has been frustrated.
The reason why this idea doesn't work is because people who know what machine learning is doing are just using open source code, and people who don't know can't do it, even if they use the API. Many smart friends are caught in this dilemma. Some people have been acquired by large companies to enhance the machine learning team (IBM's Alchemy API, Intel's Saffron, Salesforce's Metamind, etc.). However, the hot money brought about by the machine learning model behind the creation of API functionality still attracts a large number of developers.
Amazon, Google and Microsoft are all trying to sell the MLaaS layer as part of their cloud strategy. I haven't seen startups or big companies using these APIs, but I've seen a lot of artificial intelligence applications, so it's unlikely that the sample size I've observed is too small.
Services from large cloud service providers will end up in the same way as startups, as their situation swings this year.
Cloud service providers will leave these services, but will not make big money on this, MLaaS startups will start to meet the end of the year, because the growth of the swing does not understand, and no appetite has doubled.
There is a Lvery practice problem here; the MLaaS solution has no customer segmentation – they are both segmented for customers with capabilities (machine learning capabilities) and customer segments with no machine learning capabilities.
In terms of matching subdivisions: You need machine learners to help build a true product machine learning model, because it is difficult to train and debug these things well, and it also requires a comprehensive understanding of theory and practice. These machine learners tend to use the same open source tools provided by MLaaS providers. Therefore, this eliminates the customer segmentation of the machine learning ability.
Customer segmentation without machine learning: Segmented customers without machine learning capabilities will not let the machine learn to run by using the API. They buy apps to solve higher-level problems. Machine learning is just one part of how to solve the problem. The technical skills to do machine learning within the company are hard to come up with, and finding "data products" talents to help you find problems and make machine learning solutions is even harder. Customer segmentation without machine learning capabilities includes any company that needs to build a strong machine learning and data product team outside of technology companies. Yes, this means that all industries in the world are a fairly large segment. If you agree with the "software is eaTIng the world" theory, it means that all companies in the world are more or less technology companies. The same is true for data companies. There is already a big gap between top technology companies and top non-tech companies. In the era of data competitiveness, this gap will be even greater.
Full-stack vertical artificial intelligence startup has development
I have been working on artificial intelligence for more than 20 years, including nearly 10 years of creating artificial intelligence startups in Silicon Valley. I am the co-founder of DCVC (a venture capital-based artificial intelligence and data company), and my experience has made me both excited and calm about the full stack vertical artificial intelligence application.
I am excited because I think that every industry will be transformed by artificial intelligence. Calmness is because the low-level task-based artificial intelligence can be commoditized faster. I think if you don't solve a high-level full-stack problem, you will fall into the commodity world of low-level artificial intelligence services, and eventually be acquired or chronically die due to lack of motivation.
Vertical Artificial Intelligence Entrepreneurship Solving full-stack industry issues requires subject-related expertise, unique data, and products that use artificial intelligence to deliver their core values. Although most machine learning talents work for consumer Internet giants and related technology companies, most of the problems are lurking in major industries outside the technology industry. If you agree with the assumption of "software is eaTIng the world," every company in every industry must become a technology company.
Focusing on vertical areas, you can find high-level consumer demand that fits well with artificial intelligence, or new requirements that are not available without artificial intelligence. These are excellent business opportunities, but they require great business understanding and expertise. In general, most artificial intelligence entrepreneurs either don't have it, are unaware of it, or are not humble enough to bring the demands of business and expertise to the full stack (mov up the stack or go full stack).
The new full-stack vertical artificial intelligence startup suddenly appeared in the fields of financial services, life sciences, medical, energy, transportation, heavy industry, agriculture, materials and so on. With the support of data and machine learning models, these startups will address high-level domain issues. Between 2017 and 2018, some of these companies will have an accounting yield of $10 million. These full-stack artificial intelligence startups may become Tesla and Solar City in the "clean technology" field.
Etching cutting die is mainly used in the die-cutting industry to cut product labels. Generally applied to trademarks, printing and packaging, various paper art, fabric creation, home beautification, office decoration, double-sided adhesive, single-sided adhesive, stickers and more.
The main characteristics of the die cuts are that it has no jointing, high precision and a higher service life than laser cutting dies. The size of the product is stable during the die-cutting process, the maintenance requirements are simple during the storage of the moulds, and long-term storage will not affect the size of the product.
We customize diverse patterns Mould products with drawings provided by customers. We are equipped with professional metal etching equipment and exposure development equipment. The accuracy of our moulds can be maintained at ±0.05mm, the material uses is high hardness 65 Mn Steel. We can guarantee that our hard mould cutting die have straight blade lines, consistent width and height, as well as the blade height and bottom thickness accuracy also meet the customer's requirements.
Hard Mould Cutting Die,Diverse Patterns Mould, High Hardness 65 Mn Steel,Die cuts
SHAOXING HUALI ELECTRONICS CO., LTD. , https://www.cnsxhuali.com