Shulou(Shulou.com)06/03 Report--

If someone asks you what is AI? What would you say? You may think of the scenes in the movies. Star Wars and Avengers sci-fi movies provide us with a lot of imagination about AI, but those are unrealized AI robots. In real life, AI is widely used, including smart phones, search engines, intelligent games, self-driving cars and other fields, but these are the applications of weak intelligence AI.

Now AI represents the AI with strong intelligence. It is an intelligent entity that can learn deeply from vast amounts of empirical data and can make different responses according to different situations.

AI may be a new field in the cognition of most people, but its history can be traced back to the machine intelligence proposed in the last century. Although AI technology has appeared for a long time, with the development of time and the development of new software and hardware, it began to appear in the public view in recent years, and continues to develop with a strong trend.

AI technology is not unfamiliar to the public, you may have heard the names of other technologies, such as machine learning, neural network, language processing, image recognition, etc., in fact, these are a subset of AI technology, AI technology includes a very wide range, the following is the author's summary of AI technology.

Machine learning

Machine learning is somewhat similar to human schooling. When an AI program acquires a large amount of empirical data and wants it to be transformed into available information, it also needs to reorganize the classification in the existing knowledge structure and constantly repeat experiments, just as children need to make judgments and actions in the real environment by learning and imitating the behavior of adults in school.

Whether the AI program can respond correctly depends on the information set. If the program collects separate, limited, biased information sets, it is difficult to get the correct response. The difficulty of collecting the information set lies in the scope of use of the AI program users. If the AI program is created for minority users, it uses a smaller information set, but most AI programs need to collect a wide range of information and conduct a large number of repeated experiments to get the correct response to different instructions to meet the needs of most users.

Deep Learning and Neural Network

Both are a subset of machine learning, and deep learning focuses on trying to mimic the way the human brain works. Although we do not fully understand how the human brain works, now the deep learning ability of AI programs can simulate the human brain, spontaneously carry on deep learning to huge amounts of data, and construct response mechanisms in different scenarios.

Neural network is the concrete realization of deep learning. By completing different tasks, multiple nodes in the AI program share the experience of completing the task with other nodes in the network. If the task of one node is completed successfully, it means that other nodes have the experience of completing the task. And the more experience between the nodes, the more abundant the information set, and the AI program can mine more available information in the information set. It is transformed into different reaction mechanisms.

Cognitive Computing and Natural language processing (NLP)

Cognitive computing is a new computing model, including information analysis, natural language processing (NLP) and other technologies. Cognitive computing is the key technology system to achieve AI. Cognitive computing represents the ability to think and perceive transactions, and can reason, learn and improve data.

Natural language processing is a science that integrates linguistics, computer science and mathematics. NLP distinguishes written language from spoken language, understands the text content and context contained in the language, judges the ambiguity of individual words and distinguishes the speaker's intention. When the AI program deals with specific problems, NLP helps the AI program correctly analyze the speaker's instructions and output appropriate reply sentences, while the cognitive system is responsible for exploring more natural ways to interact with the interaction and helping the AI program to improve its humanized function.

Computer vision.

Computer vision uses cameras and computers instead of human eyes to help AI programs correctly identify targets, including image acquisition, preprocessing (smooth denoising, improving contrast, adjusting scale), feature extraction, detection and segmentation, and so on. Computer is used to process, analyze and understand images to identify targets and objects of different patterns. IPhone X face recognition is the application of this technology.

In a computer environment, "vision" also includes images and other types of visual input that our eyes do not normally see or recognize, but machines can recognize them, such as electromagnetic waves, X-rays and infrared rays.

Tool recommendation

In terms of function library recommendations, each technical section has some specific differences. Many libraries in the AI domain are aimed at Python developers, but the libraries mentioned below can support other languages.

Keras: advanced neural network Python library, which can be embedded in other deep learning libraries to simplify the experimental model.

MXNet:Amazon deep learning library, supporting a variety of programming languages and deployment methods.

Deeplearning4j: JVM-based deep learning library, built-in visual notebook for experimental operation.

Spark MLib: suitable for users who use Spark for data stream transmission.

OpenCV: a widely used library that supports computer vision.

SimpleCV: very similar to OpenCV.

NLTK: a Python library for processing and understanding natural languages.

OpenNLP: for JVM users.

Finally, it should be pointed out separately that as the AI program continues to develop, it will assume more diverse roles in social life, and we will increasingly rely on AI to undertake more work or tasks. although we do not believe in the robot killers in sci-fi movies, and the current AI has not achieved a strong intelligence era, we still need to be wary of the controllable challenges brought by AI.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.