We often see many powerful artificial intelligences:
In science fiction movies,
it is the terminator from the future.
In the real world,
it is AlphaGo, which easily defeated a nine-dan Go player.
We can’t help but ask,
where does the immense intelligence potential of computers come from?
How does it become increasingly “smart”?
The answer comes from machine learning.
Machine learning is a branch of artificial intelligence that studies human learning behavior. It draws on perspectives from cognitive science, biology, philosophy, statistics, information theory, cybernetics, computational complexity, and other disciplines or theories. Through basic methods such as induction, generalization, specialization, and analogy, it explores the laws of human cognition and learning processes, establishing various algorithms that can be automatically improved through experience. This enables computer systems to learn specific knowledge and skills autonomously, creating task-oriented learning systems with specific applications.
Human-like “Machine Learning” (Video Source: Science Popularization China)
Machine learning aims to use computers as tools to realistically and timely simulate human learning methods. It can categorize existing content into knowledge structures and be widely applied to solve complex problems in engineering applications and scientific fields.
New Methods for Scientific Training
Currently, machine learning provides new methods and pathways for scientific training. In competitive events like cross-country skiing and speed skating, where speed is paramount, the athlete’s velocity is the ultimate manifestation of technical movement. The research team from Beijing Sport University integrates multi-layered data, including different technical types of skiers, technical action features in different terrains during competitions, and competition results, to analyze the key technical action features that are crucial for competition speed. Based on the key technical features, they utilize deep learning methods to construct neural network models to predict skiing speed.
Identifying Potential Online Offenders
According to a report by Science and Technology Daily on December 13, 2021, computer researchers from Tokushima University in Japan, in collaboration with the large Japanese network company Cyber Agent, published a paper in Human Behavior Computing. They used machine learning methods to analyze user data from a social game under Cyber Agent and were able to accurately identify potential online offenders and predict the approximate timing of illegal behavior based solely on basic information such as chat frequency, chat partners, and chat times, without monitoring chat content.
Predicting the Synthesis of Complex New Materials
A study published on December 22, 2021, in Science Advances by researchers from Northwestern University and Toyota Research Institute successfully applied machine learning to guide the synthesis of new nanomaterials, eliminating barriers related to material discovery. This trained algorithm can accurately predict important catalysts for clean energy, chemistry, and automotive industries by defining datasets.
Predicting New Drugs Not Yet on the Market
A breakthrough in computational biology published on November 15, 2021, in Nature Machine Intelligence involved a research team, including members from the University of British Columbia in Canada, developing an automated, generative machine learning method that can determine the chemical structure of unknown new psychoactive substances (also known as synthetic drugs) using only mass spectrometry. Understanding these structures can help forensic laboratories identify suspected synthetic drugs more quickly.
Estimating Brain Age
According to a report by Beijing Youth Daily on November 18, 2021, a study published by scholars from the Australian National University showed that an increase in a person’s blood pressure, even within the normal range, can accelerate brain aging. In this study, researchers estimated brain age using machine learning methods. The results found that if a person’s blood pressure is well controlled, their brain is at least half a year younger than their actual age. In contrast, those with elevated blood pressure tend to have older and less healthy brains, increasing the risk of stroke and dementia. Even slight increases in blood pressure within the normal range can lead to brain aging and increased risks of other health issues.
Soft X-ray Radiation Distribution of the Corona
On July 21, 2021, a reporter from Science and Technology Daily reported that researchers from the Yunnan Observatory of the Chinese Academy of Sciences successfully predicted soft X-ray radiation in the corona using machine learning methods for the first time. They employed a deep learning method known as artificial intelligence convolutional neural networks to statistically analyze paired data and establish an observational mapping model. The study showed that this model can construct soft X-ray data consistent with real observations, making it more convenient and accurate than traditional methods that rely on extreme ultraviolet observations to infer soft X-ray measurements.
Algorithm Confirms 50 Exoplanets
According to a report by Phys.org on August 25, 2020, British scientists developed a new machine learning algorithm that confirmed 50 exoplanets. This marks the first time astronomers used machine learning technology to analyze potential planet samples and determine which ones are real and which are “fake” or false positives, thereby calculating the probability of each candidate planet being a true planet.
Three-Year Action Plan for New Infrastructure Construction of the Internet of Things (2021-2023)
In November 2021, the “Three-Year Action Plan for New Infrastructure Construction of the Internet of Things (2021-2023)” was issued, proposing to conduct research on artificial intelligence technologies such as voice recognition, video recognition, machine learning, object operation mechanism models, and knowledge graphs, enriching the interactive means of perception terminals and enhancing the knowledge model accumulation and specialization level upgrade in IoT services.
“14th Five-Year Plan” for the Development of Chinese Film
In November 2021, the “14th Five-Year Plan for the Development of Chinese Film” was officially released. The plan proposed to establish a national high-tech research laboratory for film, relying on national-level film research capabilities, certified by the state, focusing on research on new generation information communication technologies and intelligent science technologies such as cloud computing, big data, 5G, VR, artificial intelligence, machine learning, deep learning, trusted computing, and blockchain in the overall solution for information construction, cloudization, and intelligent upgrading of the entire film industry chain.
Academician of the Chinese Academy of Engineering
Wang Yaonan
The future development direction of artificial intelligence in robots is very important, with many key technologies determining the future of robots. Among them, memory technology, perception technology, action planning, and machine learning are all rapidly developing in artificial intelligence in recent years, and can be completely transplanted into robots. The future robots will definitely have a complete, intelligent, autonomous, and networked control system.
Deputy Dean of the School of Journalism and Communication, Chinese Academy of Social Sciences
Huang Chuxin
In the future, cutting-edge technologies will provide greater imaginative space for the construction of network integrity in our country. Artificial intelligence technology will continuously improve information supervision and screening mechanisms through machine learning, ensuring the procedural and standardized construction of network integrity. Blockchain technology, with its decentralized and secure characteristics, will build a network integrity system, activating the endogenous motivation for industry development. As quantum communication technology further develops, the construction of network integrity will surely advance to a new stage of quantum encrypted transmission, helping our digital economy to move steadily forward.
Professor of Economics and Management, Wuhan University
Zhuang Ziyin
Machine learning technologies, such as deep learning, are feasible methods for leveraging the value of big data. Machine learning is driven by big data sources, suitable for rapidly changing large and complex datasets, and can be further improved through cloud computing and edge computing infrastructure. Therefore, merging big data and machine learning is beneficial for organizations to enhance data value and expand their big data application analysis capabilities, while improving the performance of big data applications can further increase data commercial value.