Zhejiang Lab

What We Are Innovating ①: Go Deeper for Integration and Broader for Research

At present, as Artificial Intelligence (AI) is no longer a strange word, AI+ is deeply affecting our work and lives. Applications such as voice recognition for smart homes, AI-aided diagnosis for medical images, and collaborative robots for manufacturing optimization are not only made possible by the development of voice recognition, machine vision, machine learning, and other AI technologies, but also by massive cross-border and innovative studies. For the prosperity of AI, we must take AI technologies into real-life application. With a philosophy of application-oriented research, science professionals at Zhejiang Lab have spontaneously organized various academic exchange events such as the Zhejiang Lab Seminar, PI (Principal Investigator) Insights, and the Zhejiang Lab Salon to promote regular cross-center and inter-disciplinary integration and innovation.

During a recent PI Insights event, Doctor Li Taihao, Assistant Head of the Lab’s Research Center for Current Artificial Intelligence Theories, shared his ideas on intelligent education, and discussed AI+ Education with colleagues at the Lab. He pointed out that, “For the development of education, it is a prevailing tendency and inevitable trend to deeply integrate AI with education and build a new education ecology. In the future, the development of intelligent education will be mainly focused on areas such as theoretical systems for Intelligent Education, deep mining and application of education big data, personas and knowledge graphs, the management of learning processes, and emotion recognition and psychological counseling.”

As an expert in affective computing engineering, Li has long been interested in the application of AI in educational psychology. He believes that, through multimodal databases derived from the analysis of multivariate information such as facial expressions, voices, texts, and physiological signals, AI can enable machines to perform affective computing in complex environments. It can thus help teachers to better understand students' psychology and behaviors, allowing them to predict students’ mental states and make effective intervention. During the event, Li’s sharing greatly aroused participants’ interest in the research of intelligent education, leading to a lively discussion on related algorithms.

During an internal discussion of the Lab’s organizational culture, Wang Yansong, a senior research fellow from the Research Center for Industrial Internet, proposed to organize cross-center academic exchange activities on a regular basis. “In addition to exchanging and inspiring ideas, we also need to integrate resources to facilitate our research,” said Wang. Owing to the efforts headed by the Department of Research Development, his idea was quickly put into practice with the opening of a weekly event named the Zhejiang Lab Seminar. On November 6, through a presentation entitled “The Path of Digital Transformation and Industrial Internet”, Wang shared his understanding and ideas about the integrated development of the Industrial Internet, and technologies such as AI, 5G, and Digital Twins. He also introduced the application of mimic defense in the Industrial Internet based on the Center’s ongoing research projects. “The security of the Industrial Internet is no longer a problem purely belonging to the field of industrial control or information security. We need to combine multiple fields,” said Wang. To further explain this idea, he took the research of big data in the Industrial Internet as an example. It not only involves intelligent perception to realize more accurate data acquisition through sensors, but also involves intelligent computing to analyze and process multi-source heterogeneous data through better AI algorithms. How to achieve “small data, big intelligence” is the main challenge for AI in the Industrial Internet. "These problems cannot be solved by our center alone, so I am very much looking forward to engaging the teams from the Intelligent Perception Research Institute and the Artificial Intelligence Research Institute through our communication in the Zhejiang Lab Seminar, so that we can work together to overcome these problems,” said Wang.

The seminar has not only become a concrete channel for PIs to promote integrated innovation, but also a sharing and collaborative platform for young researchers. After the completion of her doctoral study in Peking University, Ma Jing joined the Lab’s Research Center for Connected Healthcare Big Data, with a research interest in omics- and clinic-based intelligent healthcare. She signed up for the seminar, and introduced the challenges and research prospects for precision medicine through a presentation entitled “Biomedical Big Data Mining and Its Clinical Application Research” . She believes that the key to the use of omics data in clinical application is the standardization and integration of systems. As AI has provided new solutions for the in-depth transformation of big data in precision medicine, instead of manual interpretation, machines can be used to comprehend complex and high-dimensional omics and clinical data, so as to assist with the early diagnosis and targeted therapy of diseases. “During the 13th CASP (Critical Assessment of Protein Structure Prediction) Competition in 2018, all of the top five algorithms such as AlphaFold integrated deep learning and traditional algorithms. Marking new breakthroughs in this field, they can facilitate important studies based on protein structure, such as the development of targeted therapy drugs,” said Ma. With detailed case analysis, her sharing provided a clear general picture of AI+ Healthcare for colleagues in other research fields. On September 23rd, Doctor Ming Yao from the Hong Kong University of Science and Technology shared his research progress on the interpretability and visualization of machine learning. For Zhang Ying, a researcher from the Research Center for Connected Healthcare Big Data, his sharing was enlightening. “I am currently engaged in the front-end development for the multi-center intelligent medical information platform. This presentation inspired me a lot. The visualization of machine learning can directly present the process of medical data processing on our platform, so that the user experience of medical practitioners and researchers can be further improved,” said Zhang. She also added that, “The communication between different disciplines and majors has given me new ideas about my own research. It is conducive to the stimulation and cultivation of innovative thinking. I look forward to seeing more cross-disciplinary ideas sparking."

At Zhejiang Lab, in addition to regular offline academic exchange events, there are also online research discussions taking place anytime. On October 10th, a viaduct on National Highway 312 in Wuxi suddenly collapsed, killing three people, and injuring two. The accident triggered a discussion in the WeChat group for the Lab’s researchers. “The collapse of the viaduct revealed the lack of health monitoring and early warning mechanisms for urban infrastructure. We can consider using fiber optic sensors for integrated monitoring and early warning,” Doctor Yan Guofeng from the Center for Super Perception claimed, being the first one in the WeChat group to express ideas about this accident. Professor Zhang Ji from the Research Center for AI Algorithms and Platforms, pointed out that an intelligent perception and early warning system for overweight vehicles could be helpful. He also introduced in detail the four modules involved in the system, including the intelligent perception module, the IoT (Internet of Things) communication module, the data analysis and intelligent computing module, and the emergency response module.

This technical discussion did not end up in the WeChat group. Instead, it was taken to a larger range of online forums. On October 14, Doctor Yan was invited to an online interview held by UWell. With technical solutions provided by the brainstorming of the Lab’s researchers, he responded to the media concern of “Are There Any Black Technologies for Overload Warning”. The forum engaged 42 media agencies, and relevant news reports all employed Doctor Yan’s expert opinions, such as “How Technology Can Help Treat Vehicle Overload” by Jiemian News, “Information Island Traps Truck Overload Warning Systems” by China Quality Daily, “Intelligent Perception in the 5G Era makes Truck Overload Treatment No Longer a Problem” by Beijing Youth, and “Unsolved Problems Left by Frequent Bridge Safety Accidents” by The Democracy and Law. From internal discussions to public communication, researchers at the Lab are endeavoring to fulfill their duties and obligations as science professionals.

In addition, researchers from multiple research centers have decided to meet for in-depth studies on the health monitoring and early warning mechanisms for urban infrastructure. Yang Wenjiao from the Research Center for Industrial Internet offered to share previous experience in the design and implementation of health monitoring projects for bridges, tunnels, and ramps. Yu Qizhi from the Research Center for AI Algorithms and Platforms expressed the intention to work with colleagues at the Lab for cross-border cooperation and achieve more socially influential outcomes for the bridge monitoring project he was currently applying for. “We need more lateral communication and cooperation, making inter-disciplinary integration and innovation a feature and a strength of the Lab.” This has become a consensus among us Lab constituents, after the discussion on this social issue.

The real-life application of forefront technologies is not a simple job, but it is one of the goals of Zhejiang Lab. We expect that the ongoing efforts for inter- disciplinary integration and innovation will take root in the Lab, facilitating the breakthroughs of original innovation.