First, understand the development of AI.
1. What is artificial intelligence?
Artificial intelligence includes both artificial and intelligent aspects. Artificial means synthetic and artificial, and intelligence can be divided into thinking school, knowledge threshold school and evolution school. Artificial intelligence is an interdisciplinary subject between natural science and social science, which integrates information, logic, thinking, biology, psychology, computer, electronics, language robots and other disciplines. The basic discipline is mathematics and the guiding discipline is philosophy. Can be defined in a narrow sense and a broad sense. In a narrow sense, artificial intelligence is a branch of computer science, which is a science and technology that simulates or realizes intelligence by computer and studies how to make machines intelligent (especially how to realize or reproduce human intelligence on computers). Artificial intelligence in a broad sense is a comprehensive subject that studies and develops theories, methods, technologies and application systems that simulate, extend and expand the intelligence of human beings and other animals, and develop various machine intelligence and intelligent machines.
Applications: intelligent express service, intelligent planning travel plan, topic photo analysis.
2. What are the types and genres of artificial intelligence?
According to whether artificial intelligence can really realize reasoning, thinking and solving problems, artificial intelligence can be divided into weak artificial intelligence and strong artificial intelligence.
Weak artificial intelligence: refers to an intelligent machine that can't really reason and solve problems, doesn't really have intelligence and independent consciousness, and only focuses on completing a specific task. Such as search engines and smart phones.
Strong artificial intelligence: refers to an intelligent machine that can really think, with perception and self-awareness. Can be divided into human and non-human.
Artificial intelligence can be divided into three schools: symbolism (IBM Deep Blue Chess Competition), connectionism (Google kelipus Camera) and behaviorism (Google Robot Dog).
3. The development stage of artificial intelligence
The development stages of artificial intelligence can be roughly divided into formation stage, development stage and prosperity stage.
Formation period (1956- 1980): Symbolism prevailed in this period.
The word artificial intelligence first appeared at the Dartmouth conference in 1956. John mccarthy put forward the term artificial intelligence, which marked its formal birth as a research field.
In 1958, a perceptron with two-layer neural network is proposed. At that time, it was an artificial neural network, which could be used for machine learning.
In 1965, john mccarthy helped MIT quit the world's first robot system with visual sensors, which can identify and locate building blocks.
From 65438 to 0968, Shakey, a mobile robot developed by Stanford Research Institute in the United States, has certain artificial intelligence: perception, environmental modeling, behavior planning and task execution. It is the first generation robot in the world, and it opens the prelude to the research and development of the third generation robot.
1974- 1980: due to the limitation of mathematical model, biological prototype and technical conditions, artificial intelligence is at a standstill.
Development period (1980-2000): 1980, XCON expert system appeared, which can automatically select components for computer systems according to users' needs and help American digital companies save a lot of money.
1982- 1986: John Hopfield invented Hopfield network, which is a neural network that combines storage system and binary system, allowing computers to process information in a new way.
1986: BP back propagation algorithm gave birth to the development of connectionism.
1987-2000: Going into the trough again.
Prosperous period (2000-): 1997: IBM Deep Blue
In 2006, Hinton proposed a deep learning neural network to break the bottleneck of BP development.
20 1 1: Watson wins as a player in Dangerous Edge.
20 12: Convolutional neural network, Google self-driving car.
20 13: the recognition rate of deep learning algorithm is as high as 99%.
20 16: AlphaGo
20 17: AlphaGo zero, Sofia
Second, the key technologies of artificial intelligence
4. What is machine learning?
Machine learning is an interdisciplinary subject, involving statistics, system identification, approximation theory, neural network, optimization theory, computer science and brain science. Study how computers can simulate or realize human learning behavior in order to acquire new knowledge or skills. The core is to reorganize the existing knowledge structure in order to continuously improve their own performance.
5. Classification of machine learning?
(1) can be divided into traditional machine learning and deep learning according to learning methods.
Traditional machine learning: starting from some observation samples, trying to find the laws that can't be obtained through principle analysis, and realizing accurate prediction of future data behaviors or trends. The main feature is to balance the validity of learning results and the interpretability of learning models, which provides a framework for solving learning problems with limited samples. It is mainly used for pattern classification, regression analysis and probability density estimation under limited sample learning. Applications: natural language processing, speech recognition, image recognition, information retrieval, biological information.
Deep learning: it is a learning method to build a deep structure model. It is characterized by multi-layer neural network. Convolutional neural network (for spatial distribution data) and cyclic neural network (for time distribution data) are formed.
Difference: Case study: identification of dogs, cats and other animals.
Traditional machine learning needs to define corresponding facial features, such as the appearance of beard, ears, nose and mouth, so as to classify and identify objects. Deep learning will automatically find out the important features needed by this classification problem and identify the object.
(2) According to learning methods, it can be divided into supervised learning, unsupervised learning and reinforcement learning.
Supervised learning: using the limited training data set that has been marked, a model is established through certain learning strategies to realize the classification of new data. Its characteristic is to know the classification label of training samples. The feature is that there is no need to train samples and manually label data.
Unsupervised learning: using unlabeled limited data to describe the structure or law hidden in unlabeled data.
Reinforcement learning: also known as reinforcement learning, is the learning of intelligent system from environment to behavior mapping, so as to maximize the value of reinforcement signal function. The characteristic is that there is no supervisor, only one feedback message, and the feedback is delayed and not generated immediately.
(3) According to the characteristics of the algorithm, it can be divided into transfer learning, active learning and evolutionary learning.
Transfer learning: when some fields can't get enough data for model training, use the relationship obtained from data in another field to learn.
Active learning: query the most useful unlabeled samples through a certain algorithm, give them to experts for labeling, and then train the classification model with the queried samples to improve the accuracy of the model.
Evolutionary learning: The nature of the optimization problem is not high, and it only needs to be able to evaluate the quality of the solution. It is suitable for solving complex optimization problems and can also be directly used for multi-objective optimization. Evolutionary algorithms include particle swarm optimization algorithm and multi-objective optimization algorithm.
6. What is big data?
Big data refers to dynamic data collection including collection, preservation, management and analysis. It is characterized by scale, high speed, diversification, value and authenticity.
Educational application: educational data mining and learning analysis
Educational data mining is to quantify, analyze and model the learning behavior and process, and analyze all the data generated in the teaching and learning process by using statistics, machine learning and data mining technologies.
Learning analysis technology is to measure, collect and analyze the data of learners and their learning environment, so as to understand and optimize the learning process and learning environment.
7. What is a knowledge map?
Knowledge map is a semantic network that maps the real world to the data world, and consists of nodes and edges. Among them, nodes represent entities or concepts in the physical world, and edges represent the attributes of entities and their relationships. There are all kinds of relationships in the real world, and knowledge map is to place them reasonably. In essence, it is a semantic network, which aims to describe concepts, entities, events and their relationships in the objective world.
From the field, it can be divided into: general knowledge map and specific domain knowledge map.
Applications: semantic search, intelligent question answering, visual decision support.
Application in the field of education: in the intelligent teaching system, the knowledge map technology is used to mine the knowledge points related to answers, so as to provide more suitable guidance and suggestions for learners.
7. What is natural language processing?
Natural language processing is a field in which computer science, artificial intelligence and linguistics pay attention to the interaction between computers and human natural language, and study various theories and methods that can realize effective communication between people and computers in natural language.
8. Natural language processing?
Includes two parts: natural language understanding and natural language generation.
9. Research field of natural language processing?
Research fields are very extensive, such as machine translation, semantic understanding, question answering system, etc. Text analysis (automatic composition evaluation system), recommendation system
10. What are the four major challenges facing natural language processing?
There are uncertainties at different levels such as morphology, syntax, semantics, pragmatics and phonetics.
New words, terms, semantics and grammar lead to the unpredictability of unknown phonetic phenomena;
Insufficient data resources make it difficult to cover complex speech phenomena;
It is difficult to describe the fuzziness and complexity of semantic knowledge with a simple mathematical model.
1 1. robotics
The first generation robots are program-controlled robots, and they can work repeatedly according to the proposed program.
The second generation robot is an adaptive robot, equipped with corresponding sensory sensors, which can change its behavior with the change of environment, but it has not yet reached the level of complete autonomy;
The third generation robot is an intelligent robot. They are equipped with various sensors, which can process the sensed information and control their behavior. They have strong adaptability, learning ability and autonomous function.
Intelligent control methods: expert control, fuzzy control, neural network control and expert hierarchical control.
12. What is cross-media intelligence?
Cross-media: Text, image, sound, video and their interactive properties will be closely mixed together.
Cross-media intelligence is the basic intelligence to realize machine cognition of the outside world. (Pan Yunhe)
13. Key technologies of cross-media intelligence?
Cross-media intelligent retrieval, cross-media analysis and reasoning, cross-media knowledge map construction, cross-media intelligent storage
14. Application of cross-media intelligence?
Smart city, medicine, education (wearable technology, brain-computer interface. Multi-modal angle)
15. Educational challenges in the intelligent age?
Challenge 1: How to cultivate talents with AI literacy?
Challenge 2: How does the educational administrator reconstruct the workflow?
Challenge 3: How do teachers cope with the impact of artificial intelligence?
Challenge 4: How do teachers apply artificial intelligence to change teaching methods?
Challenge 5: How do students use artificial intelligence technology to change their learning behaviors and ways?
Challenge 6: How to upgrade the course content in the intelligent age?
Challenge 7: How to deal with the ethical, social and security issues in artificial intelligence education?
17. What are the connotations and characteristics of AI educational application?
Intelligent education: artificial intelligence education in a narrow sense: education with artificial intelligence as its content, the purpose of which is to cultivate professionals who master machine intelligence technology to meet the needs of technological development. Intelligent education in a broad sense: education supported by intelligent technology, education to learn intelligent technology and education to promote intelligent development.
Wisdom education: with the support of information technology, develop students' wisdom ability. He emphasized the use of technology integration to build a learning environment, so that teachers can teach efficiently and students can learn individually.
Intelligent education is the education supported by technology. Intelligent technology not only makes the learning environment richer and more dexterous, but also makes the machine have human-like or even superhuman intelligence in some aspects.
Wisdom education is guided by the concept of wisdom education. The advanced concept of wisdom education determines the mode of wisdom teaching method. Different modes require teachers to have corresponding teaching skills, which can only be realized with the support of intelligent environment.
18. What are the characteristics of the application of artificial intelligence in education?
intelligentize
Artificial intelligence technology is the core technology to promote education informatization, and it has the potential to change the way of teaching and learning. In the future, there will be more and more intelligent tools in the field of education to support teaching and learning. Intelligent education will bring learners a brand-new learning experience and provide a foundation for teachers to implement high-quality teaching. With the support of educational information technology, we will build an educational information ecosystem, seamlessly integrate online learning environment with real situations, and make human-computer interaction more convenient and intelligent. Ubiquitous learning and personalized learning will become the new normal.
Man-machine cooperation
Man-machine collaborative education can give full play to the different advantages of teachers and artificial intelligence, and promote the individualized development of students. Machines are mainly responsible for repetitive, monotonous and recursive work, while teachers are responsible for creative, emotional and enlightening work.
Teaching automation
Artificial intelligence can directly apply subject knowledge, teaching method knowledge and learner knowledge to realize the automation of knowledge dissemination, so it can be used to support educational activities.
Personalized
In order to expand the teaching scale and improve the teaching efficiency, the traditional teaching organization adopts the class teaching system, which is similar to the large-scale production in factories and ignores the individual differences among students. With the advent of the era of artificial intelligence, it is possible for teaching organizations to tend to personalized education. Artificial intelligence can analyze each student's process learning data, accurately identify their knowledge level, learning needs and personal hobbies, and build a learner model, thus realizing the push of personalized resources, learning paths and learning services. This means the end of the era of mass production education and the beginning of personalized education.
Interdisciplinary integration
Artificial intelligence involves many disciplines, and the teaching of a single discipline can no longer meet the needs of social development, so interdisciplinary integration teaching is highly respected. In order to adapt to the individualized development of students, future education should cultivate students' diversified and all-round development. Project practice takes artificial intelligence as the core, provides real problem situations, and focuses on stimulating, cultivating and improving students' computational thinking, innovative thinking and meta-cognition.