With the rapid development of artificial intelligence technology, all walks of life have been profoundly affected, especially the popularization of artificial intelligence education in primary and secondary schools, which has become an important course that cannot be ignored today. The emergence of artificial intelligence technologies such as ChatGPT allows us to see that artificial intelligence will become one of the important driving forces to promote the leapfrog development of science and technology and the overall leap of productivity in the future. In this context, primary and secondary education has taken artificial intelligence as an important learning content, and scientific and technological innovation activities and popular science activities as representatives of interdisciplinary learning activities. In the future, artificial intelligence will become one of the most important learning and work assistants for teenagers. Therefore, scientific and technological innovation activities that integrate artificial intelligence technology will become an indispensable part of artificial intelligence education in the future.
Artificial intelligence is a new research tool for young people's scientific and technological innovation activities
With the development of artificial intelligence, machine learning is increasingly becoming a core tool for scientific discovery. Scientists believe that "machine conjecture + scientific intelligence" will become the new paradigm of scientific discovery. In youth scientific and technological innovation, focusing on cultivating the ability to solve problems using artificial intelligence can help young people form a new understanding of future scientific research. Artificial intelligence has broad potential in youth scientific and technological innovation and can play three roles.
1. Liberate manpower and assist in handling complex and cumbersome data processing tasks
Data exploration has attracted much attention in youth science and technology innovation activities, but data collection and collation is one of the most tedious tasks. The use of artificial intelligence can liberate manpower, realize automatic identification, intelligent processing, etc., and better reflect the important value of artificial intelligence.
For example, cameras are used to achieve automatic recognition in scientific observation experiments, cameras are used to count quantities in seed germination experiments, and a large amount of unstructured data is intelligently processed.
Workflow diagram for the Scientific Observation Aid
Wenzhou Middle School students wrote an information system that simultaneously records image and sensor data. This "scientific observation aid" can be run on the "row empty board" or "virtual valley", using a camera to collect footage, and record soil information through soil moisture sensors. When the camera image changes (e.g. flowering, fruiting, pests, etc.), or the value of the sensor changes, the image is automatically saved and the corresponding information is marked on the image. Among them, to determine whether to flowering and fruiting with pests, the target recognition artificial intelligence model trained by MMEdu is used. With the help of such systems, people can be freed from tedious and boring observations.
Second, find a different way to solve mathematical problems by collecting data
Students will encounter difficult problems that need to be solved mathematically in scientific and technological innovation activities, and neural networks can be used to fit the relationship between multiple physical quantities, which is helpful to cultivate data awareness and computational thinking.
For example, by using deep learning to simulate mathematical formulas, students who have not studied trigonometric functions in upper elementary school can complete complex obstacle angle calculations. By using the neural network library BaseNN, complex functions can be effectively simulated with only a few layers of simple fully connected neural networks and a small amount of training, which greatly reduces the difficulty and solves mathematical problems that are difficult for elementary school students to understand.
3. Pattern recognition, simplifying data programming for complex scenarios
In youth science and technology innovation activities, the difficulty of programming using multiple sensors has increased exponentially, especially in more complex scenarios such as time and space changes. However, the use of machine learning can greatly reduce the difficulty of programming.
For example, Wenzhou Middle School students used machine learning to develop a magic wand that can recognize different gestures and perform different functions, such as lighting a torch and opening a curtain. The swing of the magic wand is actually a physical quantity that changes the acceleration. They chose the control board, repeated the acceleration data labeled by gestures, and trained a model with a neural network to recognize different gestures. After testing, they achieved a good recognition effect.
Artificial intelligence is a new research object for youth scientific and technological innovation activities
Unlike other classical computer algorithms, the training of AI models is particularly closely correlated with datasets. When faced with a real problem situation, even with the support of the latest SOTA model (SOTA is short for state-of-the-art, which refers to the best and most advanced model currently performing in a research task), many parameters need to be fine-tuned, so AI experts will jokingly call training an AI model "alchemy". To train an excellent model, you have to do a lot of experiments, and even need to detail the classical model, which is worth constantly studying.
1. Joint work of different algorithm models
When faced with complex recognition tasks, the practice of joint work using multiple different SOTA models can achieve good results. For example, recognizing a specific gesture in an image can use a combination of a human recognition model and an image classification model, or recognize both a human body and a gesture using multi-object detection. This "clumsy" but effective approach is also applicable to young people's scientific and technological innovation activities, because artificial intelligence is a new research object worth exploring.
2. Artificial intelligence algorithm design in low computing power environment
In primary and secondary schools, the lack of computing power is a common phenomenon. How to train a model in limited computing power, or achieve fast inference in limited computing power, is worth in-depth study. Relatively speaking, some artificial intelligence application platforms with built-in models run faster, such as OpenCV recognition of faces, MediaPipe recognition of face landmarks, recognition of gestures, faster than our self-trained network, making full use of the characteristics of rapid recognition of artificial intelligence application platforms, and further training models on the basis of its recognition results, which is a feasible low computing power solution.
Artificial intelligence is a new weapon for teenagers to solve engineering problems
The Science Curriculum Standards for Compulsory Education (2022 Edition) states that "inquiry practice mainly refers to the ability of scientific inquiry, technical and engineering practice and independent learning formed in the process of understanding and exploring nature, acquiring scientific knowledge, solving scientific problems, and technical and engineering practice." "It can be seen that engineering problems are an important direction for scientific and technological innovation activities. Designing various maker works with the help of open-source hardware such as Arduino and control board in the maker space is one of the main forms of youth scientific and technological innovation activities. Now, artificial intelligence provides teenagers with new weapons and can solve more problems.
Taking the design of a "smile robot" that greets children with smiling faces as an example, in addition to training and deploying models that recognize smiling expressions, several related tools are required: tools that acquire camera images in real time, such as OpenCV; Hardware and software tools capable of driving servos, such as pingpong and control boards; Mini PCs capable of deploying artificial intelligence models, such as row blank boards; Tools that synthesize speech, such as Pyttsx.
With the popularization of AI education, more and more regions and schools are participating in AI technology innovation activities. The Ministry of Education and Shanghai Artificial Intelligence Laboratory have also launched a practical research project on artificial intelligence education in primary and secondary schools and a new generation of artificial intelligence teacher growth camps, making artificial intelligence science and technology innovation activities in primary and secondary schools "fly" into the classrooms of more ordinary schools, making artificial intelligence an indispensable element in primary and secondary school scientific and technological innovation activities.
Note: The text views are excerpted from the article entitled "[Artificial Intelligence Education in the Context of New Technology] When Artificial Intelligence Meets Youth Science and Technology Innovation Activities" published by teacher Xie Zuoru (Wenzhou Middle School, Zhejiang Province) in the public account of "China Science and Technology Education"