In 2017, China's artificial intelligence startups accounted for 48% of the world's total, and surpassed the United States for the first time. In 2018, according to CBInsights, the us will still lead the world in terms of the number of artificial intelligence start-ups and total equity deals, but it is losing its global dominance.
Trend 1 China potential explosion, shake the United States dominant position
Two of China's biggest achievements are facial recognition and smart chips. The former has benefited from strong government support in recent years, while the latter is a direct challenge to the traditionally strong American chip.
Megvii, a unicorn, is a major contributor to facial recognition. The Megvii (sunshine insurance group) by the Chinese insurance company, government agency (Russian federation - China investment group), and the giant foxconn, ants, financial support, now has 1.3 billion face data records.
China has invested heavily in face recognition.
Alibaba group, the company's investors (by ants financial) and cooperating foxconn in China in 2016, hangzhou is the "brain" projects, using artificial intelligence to analyze surveillance camera data.
"I think there has been a lot of change from the past to the next year," said Dr. Wang jian, chairman of the ali technology committee, in the city's brain. If the last time we really did a huge experiment, the next year we're going to turn this experiment into the most basic thing in a city. One of our goals in hangzhou this year is to transform the city's brain from past trials into one that covers the entire city of hangzhou.
In July 2017, the Chinese government said it would work with the us in 2020 and become a world leader by 2030. Cambricon, a Chinese company, promises to produce one billion processing units over the next three years and is developing chips dedicated to deep learning.
In addition to domestic research and development, China's leading technology giants tencent, baidu and jd.com have increased their overseas investment. Recently, baidu and jd.com have invested in ZestFinance, and tencent has invested in ObEN. In 2018, it will also be a powerful bargaining chip for China's chips.
Trend 2 manufacturing workers or the most severe unemployment wave in history
The Chinese T-shirt manufacturer tian yuan clothing has signed a memorandum of understanding with the us state of Arkansas to enable 400 "workers" at its new clothing factory in Arkansas. It is worth mentioning that the 400 sewing robots developed by the Georgia startup SoftWear Automation. In this cooperation, all the work is done by robots, and human workers are only responsible for the maintenance and operation of robots and other high-end work.
The efficiency is greatly increased, the cost is greatly reduced, there is no complaint, there is no strike. It is not hard to imagine that in the future it will become the norm of manufacturing. Thus, in 2018, ordinary workers or the most severe unemployment wave in history.
Achieving complete production with zero labor is the ultimate goal. The same is true of amazon's unmanned warehouse concept. Amazon now employs more than 100,000 robots in warehouses around the world, and most of its work is done.
As a result, thousands of new high-tech jobs have emerged as robots replace ordinary workers. In the amazon, human workers are focused on detailed work, such as the selection of goods and the distribution of orders. It can be seen that the relationship between robots and humans is more complementary and win-win. The simple "alternative threat theory" is nonsense.
Trend 3 artificial intelligence everywhere, machine learning omnipotent
Artificial intelligence is everywhere in 2018. Or, more precisely, machine learning will be everywhere. In the view of CBInsights, the technology is almost "omnipotent" and will create an infinite number of possibilities in 2018.
"Are you a vegetarian, gluten-free or allergic to soy?" American Prose wants to use machines to customize hair products and has raised $7.57 million from well-known venture capital firms. More unexpectedly, machine learning has dabbed in the field of cannabis technology. DeepGreen now USES computer vision to identify the sex and health of marijuana plants. In addition, Weedguide has raised $1.7 million and plans to use artificial intelligence to personalize weed recommendations.
The disruption is far from enough. In 2018, based on this technology, Britain's IntelligentX is expected to launch the world's first AI brewing beer. Russia's DeepFish is committed to using neural networks to identify fish in radar images. Hoofstep, Sweden's Hoofstep, has raised venture capital funds and plans to conduct in-depth analysis of horses.
Trend 4 network security and the traditional defense world gradually converged
During the cold war, governments repeatedly talked about the "missile gap" and saw it as the key to success. Today, the battlefield is shifting to "data centers". In particular, with the development of artificial intelligence technology, the gap between governments in the network capacity is emerging, and the network security and the traditional defense world are gradually converging.
Artificial intelligence has a natural advantage in defense. Because cyber attacks are evolving, the defense process often requires malicious software that is previously unknown. Artificial intelligence, on the other hand, can stand out by its massive computing power, quickly sifting through millions of events to detect signs of anomalies, risks and future threats.
In fact, in 2014, amazon created a custom cloud computing service for the CIA to meet the strict compliance and regulatory requirements of sensitive data. In the fourth quarter of 2014, AWS opened such tools to other government customers outside the intelligence community. Meanwhile, amazon has bought two artificial intelligence network security companies, Harvest. Ai and Sqrrl, to protect sensitive data in the cloud. Over the next few years, big brother, amazon, and many artificial intelligence start-ups promised to be the backbone of the new cybersecurity effort.
According to statistics, in the past five years, 134 start-up companies have received financing of $3.655 billion. Last year, about 34 companies made ipos, joining the markets of Cyber eason, CrowdStrike, Cylance and Tanium. Each company is valued at more than $900 million.
Trend 5 voice recognition technology spring
Amazon Echo and Google Home are the stars of the consumer electronics show in 2018. The Internet of things that people can think of is integrated into it, and there are few "missing fish". Industry insiders predict that the voice recognition spring is coming.
It's worth noting that while amazon has a lead in voice computing, it lags behind in terms of language support. Amazon announced last quarter that it would sell its Alexa based speakers in about 80 countries. But the downside is that it wants global users to interact with it in English, German or Japanese. In terms of language, rival GuGe has obvious advantages. Its intelligent assistant can provide English, French, German, Italian, Japanese, Korean, Spanish and Portuguese versions. Its speech recognition feature supports 119 languages.
In addition to the two giants, samsung is now developing its own voice assistant, Bixby. We hope that all products can be connected via Internet and realize the comprehensive intelligence of Bixby by 2020. In 2017, all LG devices support wi-fi connection, and more than 80 products are integrated with GuGe Home. In China, alibaba reported that more than one million copies of Tmall have been sold since its official launch in July 2017.
In 2018, the voice market of non-english speaking countries will be a "fat meat". Who is the master of the ups and downs, also ask each elder brother to use product to speak.
Trend 6 artificial intelligence "marginalization"
In 2017, artificial intelligence is gradually brought to the edge of application, and smaller devices and sensors will be running closer to the computing network. In other words, artificial intelligence may be "hidden in your headphones" rather than on a cloud or a smartphone.
For example, apple released the A11 chip, which includes a "neural engine" for the iPhone 8 and iPhone X. Apple says it can perform machine learning tasks at a speed of 600B per second and support many new features such as FaceID. Specifically, the "neural engine" can scan a user's face with invisible light without having to upload or store any user data in the cloud. In addition, Intel has released a visual processing chip called Myriad X (originally developed by Movidius, which was acquired by Intel in 2016). Intel says it can use the deep learning technology of smartphones for baby monitors and unmanned aerial vehicles.
However, the marginalization of artificial intelligence is far more than that. In many popular fields, such as smart home and autonomous driving, it has its own figure.
Trend 7 $1.8 billion in white? Amazon, Google monopoly intensified
According to CBInsights, an investment firm that has invested $1.8 billion in enterprise-level artificial intelligence applications in the past five years. However, with amazon and Google's incremental improvements and innovations in enterprise ai applications, the money is likely to go down the drain.
In recent years, GuGe has released Cloud AutoML. Customers can use custom data training algorithms to meet specific requirements; Amazon began selling "ai-as-a-service" and "amazon artificial intelligence" in its AWS division, working to serve small developers with zero upfront costs. In addition, Amazon has released products that work like the API, allowing any developer to access Lex (NLP), Amazon Polly (voice synthesis), and Amazon Rekognition (image analysis).
In the face of such a powerful competitor, the survival of the small business may become a false proposition.
Trend 8 the most prevalent "convolutional neural network" is subverted
Neural networks have different architectures. At present, one of the most popular types of deep learning is called the convolutional neural network. Now a new architecture, the capsule network, has emerged and is expected to overtake the convolutional neural network in multiple ways.
For a long time, the convolution neural network has been successful, but scholars generally believe that it still has defects, which may lead to a security breach. Based on this, Geoffrey Hinton, one of the pioneers of deep learning, published his research paper in 2017, introducing the concept of "capsule network". Currently, the paper is still under review and needs to be tested in practice. But the concept has caused a stir in the tech world. Industry insiders predict that, once verified, it will overturn the most popular "convolutional neural network".
It is reported that the capsule network will allow artificial intelligence to identify the image pattern with less data, and is not susceptible to the error result. For example, it can recognize that the face on the right side of the picture is rearranged, and it is no longer a face. This is what the convolutional network is not good at.
Another problem with convolutional neural networks is that they can't intelligently respond to changes in input data. For example, the user must train the convolutional neural network with images of the same object from different angles or perspectives to identify all changes. Therefore, it requires extensive training data to cover all possible changes. The capsule network is different. This technique requires less data and takes into account the relative position and direction of the object.
All of the above shows that this paper, once verified, will set off a huge storm in the field of artificial intelligence. And will it happen in 2018? We'll see.
trend 9 a million annual wage war will intensify
In 2017, the battle for artificial intelligence is officially launched. A Chinese ai unicorns recruitment notice released by hunting.com shows that senior machine learning researchers pay $567,000 to $624,000 a year, and machine learning specialists earn between $315,000 and $410,000 a year.
According to a recent report by tencent, there are only 300,000 qualified researchers in the field of artificial intelligence, including students in related fields. Nationwide, however, the talent gap for artificial intelligence is one million or more. Therefore, in 2018 and the next few years, "getting talent" is the prerequisite and key to the development of artificial intelligence enterprises.
Not only China, but also the us tech companies are not stingy with artificial intelligence. Deepmind Technologies, which was acquired by GuGe in 2014, reported that in its 2016 financial statements, "staff costs and other related costs" amounted to 108.4 million pounds. On LinkedIn, the average salary for a group of 415 employees is as high as 252,000 pounds (about $350,000 a year).
Snapshot of financial statements of Deepmind Technologies (image from CBInsights report)
In addition, artificial intelligence researchers from large technology companies are leaving their businesses. Andrew Ng left baidu to create a $175 million AI fund. Google sensor research and development experts are leaving to serve as chief technology officer of the AI chip start-up Groq...
As a result, as the backbone of major companies is lost, the battle for talent in the global artificial intelligence company will become increasingly fierce. In 2018, the pay of artificial intelligence specialists will also hit a new high.
Trend 10 machine learning capital revel
In recent years, there has been another wave of capital mania. From big data, to cloud computing, to machine learning, "countless capital RACES".
2017 is the pinnacle of machine learning. Investors have invested more than $15.2 billion in machine-learning startups in various industries, up 141 percent from 2016. Over the course of a year, American incubators have absorbed more than 300 machine learning startups, three times more than in 2016.
In 2018, the carnival is about to end. The normalisation of machine learning will make investors particularly critical of the ai companies they invest in. As Frank Chen of 16z puts it, "within a few years, no investor would be looking for a machine learning startup. It will be identified as' a necessary tool to power start-up products'.
Like many of the previous "air vents", machine learning will soon cease to be fresh. And where will the legions of artificial intelligence startups emerge after 2016? Only a strong business model can keep it alive.
Li yanhong, chairman and CEO of baidu, said at the China IT leaders' summit: "our generation is very lucky as a whole, so we don't need to find the air. I have been in the air for 15 years since I came back to China in 2000, and I feel uncomfortable and all kinds of opportunities. If everyone is looking for a shortcut, everyone is thinking in this way, which is dangerous. The whole society should not encourage people to find this shortcut.
As for technological innovation, the amara rule has long been pointed out - we tend to overestimate the short-term impact of technology, but underestimate the long-term impact of technology. Perhaps all of the product model-oriented upstarts are the vents, but the technology itself is not wrong. Artificial intelligence is also so, if can be in the end after many foams to fall to the ground, the enterprise and the product left behind after the great wave will shine.