Since 2002, Professor Qiu has led his team in pioneering an optimisation-based HDR tone mapping algorithm using artificial neural networks and other technologies. This work has significantly advanced HDR photography techniques and applications, earning numerous international awards. Currently, his team is researching the application of large multimodal neural network models to address this challenge.
At present, Professor Qiu is leading a research team at UNNC to explore new applications of AI. "In line with 'Good Health and Well-being', one of the 17 Sustainable Development Goals of the United Nations, we have developed a deep neural network model to automatically predict the 'beautifulness score' and 'happiness score' of natural images, helping to promote mental health." Compared to other applications, there is relatively little research on using AI to improve human mental health, but this is an important direction, according to Professor Qiu.
"AI has developed to a level previously unimaginable," said Professor Qiu. "The computers, smartphones, and smartwatches we use today all rely on the von Neumann architecture." However, compared to the von Neumann architecture, which excels in computing power, the human brain has three key advantages: automatic error correction, multi-threaded parallel processing, and autonomous learning capabilities.
Can computers think like humans? Since Alan Turing, the father of AI, first introduced the idea of a "thinking machine" in the 1950s, researchers have proposed various models and algorithms.
"Artificial neural networks can be simply understood as computer models that simulate the neural networks of the brain," explained Professor Qiu. "Over the past 70 years, many scientists and researchers have made outstanding contributions to this field. Today, modern neural network systems, particularly those represented by deep learning, have reached an advanced level and are the driving force behind the current wave of AI."
"Professor Hinton is a globally renowned expert in AI, especially deep learning. His many groundbreaking works have played a foundational role in driving today's AI revolution," said Professor Qiu. "However, Professor Hopfield's name has rarely appeared in recent discussions about AI."
Why was the Nobel Prize in Physics awarded to Professor Hopfield? Professor Qiu explained that in 1982, Professor Hopfield proposed an innovative neural network model—the "Hopfield Network Model"—which could mimic human associative memory capabilities. He later implemented the model using analogue circuits, bringing AI discussions to a new peak. Professor Qiu's first related research paper applied the optimisation capabilities of the Hopfield Network Model. "The implementation of this model utilises the concept of 'energy function' from physics."