What breakthroughs have AI made?

AI sub-domains include machine learning (ML), natural language processing (NLP), deep learning (DL), robotic process automation (RPA), regression and so on. So what breakthroughs have AI made in the past year? After talking to 21 professionals, we gathered their views.


Fact

· In the past year, artificial intelligence has made many breakthroughs, especially in deep learning. For example, AlphaGo Zero can learn Go and chess by itself and play games with humans without human intervention. Taco Tron and Baidu's Deep Voice produce almost identical voices to human languages. In addition, computer vision, target detection and image segmentation have become more accurate, even in medical diagnosis and biological research can be comparable to human beings. However, technologies such as natural language processing, chat robots and text summarization have failed to achieve the desired goals.


· Artificial intelligence has existed for a long time. New and old things are progressing. It is important not to underestimate the power of public awareness. When Dark Blue defeated Gary Kasparov, the situation was different. It was only in movies that people were defeated by machines, but now it really happens, which has changed people's view to a great extent. And we have many applications that provide business value through artificial intelligence.


· Artificial intelligence is no longer considered to exist only in science fiction. Most technology companies already understand the benefits of AI to businesses. This has enabled the technology to develop rapidly in the past few months, with better profitability, and the ability of the machine to improve its learning process in real time.


· Over the past year, we have focused on building real conversational AI. Current assistants do not have the ability to deal with more complex and valuable tasks. To achieve this, artificial intelligence technology is needed. It can reasoning based on knowledge, understand incomplete or vague language through context and individualization, and make use of artificial intelligence to transcend pattern matching, so as to realize real dynamic dialogue. Just as humans communicate through gestures, gaze and other factors, we are also beginning to connect other services and virtual assistants in the system. That's why we launched cognitive arbitrator, which seamlessly connects and integrates different virtual assistants, third-party services and content through a single interface across automotive, smart home and Internet of Things (IoT) ecosystems to accomplish complex tasks and enhance user experience. Therefore, we can maximize the user's unique and individual experience, while realizing the interaction of various services between assistants. This is a win-win situation for every individual in the ecosystem of the Internet of Things, especially those who purchase products and services.


· AI and ML have moved out of the lab to more mainstream applications. Artificial intelligence is entering a new constitution, and it is just beginning. Six years ago, the title of data scientist did not exist, but now it has become very professional. Data scientists and developers have achieved faster and better tasks using artificial intelligence.


GPUs

· From 2000 to 2003, algorithmic trading was gradually adopted by all trading companies. In the past few years, machine learning has developed rapidly due to the increasing demand for applications. In situations where creativity is needed, artificial intelligence is replacing humans, because machines can make their own decisions based on new sources of signals and large amounts of data.


· Technically, over the past year, GPU-based servers have become commonplace as developers have begun to take advantage of processing power to accelerate application development. Professional processors like Google's TPU are beginning to emerge, and its competitor cloud service providers are working together to develop an open source deep learning library. In addition, it has steadily shifted from big data and point tools (such as Hadoop and Spark) to more extensive data analysis classes using artificial intelligence and neural networks. ML narrows the gap between these methods by using large and different data sets and applying algorithmic intelligence to analysis. The self-learning ability of learning algorithm is still in the primary state. Artificial intelligence plays an increasingly important role in our life. The product and service recommendation engine and image processing system have been significantly improved. Artificial intelligence has produced many new professions. The pace of innovation in this field is accelerating rapidly.


efficiency

· The concepts of AI and ML are key elements of cloud computing, but they only work when users have data. The automation program implemented by ML improves the work efficiency of employees in enterprises, and as employees become more familiar with AI tools, the degree of automation will become higher and higher. In addition, the work of simplifying data integration is on the rise. Especially, enterprises hope to get more useful information from the data. The growing concern for forecasting analysis enables enterprises to convert real-time data into action guides.


data

· Artificial intelligence is not new, but its revival is due to the ability to process the required data, as well as the speed and type of data. Information is huge and messy. It needs to use artificial intelligence to obtain useful information and data. The problem is that they don't have complete control over the data around them.

· Artificial intelligence has evolved dramatically in the past year for two main reasons: 1) All enterprises are undergoing rapid digital transformation. 2) The speed at which new business and operational data sets are introduced and how they improve people

Please read the Chinese version for details.