Machine Intelligence Company Overview

For artificial intelligence, there are two kinds of attitudes at present. One is the fear that Elon Musk and others fear that advanced AI poses a threat to humans, and Nicholas Carr is concerned that automation can rob people of their jobs and make them stupid; And now the artificial intelligence is very primitive. So what is the status of machine intelligence now? Bloomberg Beta VC Shivon zilis lasted 3 months, analyzing 2,529 of artificial intelligence, machine learning and data-related start-ups to explain to us.

What is Machine intelligence?

The so-called machine intelligence, machine learning and artificial intelligence collectively. Computers are learning how to think and read and write. are also acquiring human sensory functions, including sight, hearing, and touch, taste and smell (a little less attention to the latter three). Machine intelligence technology involves many different types of problem (classification, clustering, natural language processing, computer vision, etc.) and methods (support vector machine, depth belief network, etc.). These are included in the machine intelligence map.

Machine intelligence is inseparable from large data, which is the basis of machine learning and artificial intelligence. However, for reasons of space and focus on artificial intelligence methods, the map does not put large data into it.

Company classification

Many companies are engaged in machine intelligence, but the layout format is limited. So the machine intelligence method is considered as the key technology of the company can be included. Then the selected companies are divided into three major categories, one kind is the core technology company which concentrates on the core technology innovation of the machine intelligence; one is the application-oriented company, which can be divided into three categories: enterprise, industry-oriented and human-computer interaction (HCI). The third category is support technology, including hardware, data preparation , data collection and so on.

If you are going to open a related company, you can use this map to find the right core technology and support technology and then package it into a new industry application. While everyone wants to solve some of the intriguing problems, there are plenty of business opportunities in many less sexy industries (such as through Watson Developer Cloud, ALCHEMYAPI, etc.), so it is not necessarily a matter of keeping a close eye on the hot areas.

Layout thinking

Kevin Kelly (K.K) believes that cheap parallel computing, large datasets, and better algorithms drive the development of machine intelligence, bringing about changes to businesses, industries and humans. The application of this map is inspired by this view. As K.k said, "The next 10,000 startups ' business plans are easily predictable, doing X and then adding AI." "Sometimes even x can not, because machine intelligence itself is likely to create a new industry."

The foreground of machine intelligence is very considerable. The acquisition rate for start-ups in the field has 10%,zilis that another 10% will likely be bought by the end of 2015. There are 15 buyers, of which Google is the number one buyer in the field of machine intelligence.

Large companies have an overwhelming advantage, especially those that develop consumer products. Search (Google, Baidu), social networking (Facebook, LinkedIn, Pinterest), content (Netflix, Yahoo! ), Mobile (Apple) and E-commerce Giants (Amazon) are in a very advanced position. Because these companies have a lot of data, and can be through constant interaction with consumers, thus forming an algorithm to adjust the feedback loop, coupled with the network effect, so it is the most easy to reap the benefits of machine intelligence results from the company.

The success of these companies is facilitated by state-of-the-art personalization and referral algorithms. In the mobile new battlefield, machine intelligence is also indispensable: such as the natural Language interface (Apple Siri), visual search (Amazon's Firefly), directly provide answers rather than links to dynamic questions answered. IBM and Microsoft have also made great strides in this area, but the focus is on the knowledge representation task for large industry datasets (because of the lack of similar human-facing requirements for these companies), such as IBM's Watson, for example.

Talent Monopoly

For the past 20 years, the best people in the field of artificial intelligence are in academia. These people have invented many new machine intelligence methods, but few of them can bring commercial value. But now the complex machine intelligence methods like the depth belief network (deep belief nets) and the Hierarchical neural network (hierarchical neural receptacle) begin to solve some practical problems. And those on the ivory tower began to enter the enterprise. For example, Facebook recruited Professor Yann LeCun of New York University and Rob Fergus,google to hire the Geoffrey Hinton of the University of Toronto, while Baidu has Wunda (Andrew Ng), which are "Godfather" tasks in the field of machine intelligence. However, these people are not completely divorced from the academic relationship, a lot of time and energy to contribute to the school.

High-paying and good facilities are certainly one of the factors that attract these top academics, but the most important thing is another: data. Facebook, Google, Baidu and other large computing resources, but also monopolize the vast number of data, will inevitably attract more and more talent to join, which is the big companies to form the overwhelming advantage of the reasons.

Peace dividend

As noted above, big companies have an inherent advantage, and winning machine intelligence will be more powerful in the future. Fortunately for other companies, the core technologies developed by big companies are rapidly pouring into other areas-through the way of large companies leaving and publicly published research.

In addition, like the Big Data revolution, technology giants will also be able to make some breakthrough technology to the community, and then other people to do the application level of innovation.

Entrepreneurial opportunities

My company engages in X's deep learning

If you want your company to fire next year, you can use the words above to advertise. Of course, if you really are.

Deep learning is a popular method of machine intelligence. It may have been a bit of a hype, but companies such as Google, Facebook, Baidu and Enlitic, and so on, have done well in terms of vision and language processing.

The most exciting thing about deep learning is that, if handled properly, its automatic learning function can replace the intuition of some domain experts. In many cases, this is expected to rewrite many areas of the solution.

Talent acquisition as a business model

When we talk about big data, we often mention the shortage of data scientists. But since the previous machine intelligence was limited to academic research, machine intelligence experts are in short supply. This situation will not change quickly.

This shortage is a benefit to the founders who really understand machine intelligence. Many start-ups in this field are able to get seed-round financing, often for one reason-the purchase price of a machine smart talent is more than 5 times times the price of a general technical talent acquisition (say Deep mind per capita acquisition price of 500 to 10 million U.S. dollars). As a savvy founder, you can even get a bunch of machine smart people and then set up a company, maybe someone will buy you-well, it's a joke, but it does reflect the value of "artificial" intelligence.