DeepSeek has profoundly impacted the artificial intelligence (AI) landscape. Delivering comparable performance to tech industry leaders at a fraction of the training cost, DeepSeek has demonstrated that high-performance AI models can be achieved using fewer and less advanced graphics processing units (GPUs). It has also shown that cutting-edge AI technology can be realised through the optimisation of algorithms.

Professor Yang Hongxia, Associate Dean (Global Engagement) of the Faculty of Computer and Mathematical Sciences and Professor of the Department of Computing, is championing a Co-GenAI approach, which builds on DeepSeek’s disruption to make AI development more accessible and less dependent on massive, centralised computational resources.

A renowned AI scientist and former researcher at ByteDance and Alibaba’s DAMO Academy, Professor Yang shared her innovative approach at the recent Public Forum for Research and Innovation: DeepSeek and Beyond, hosted by the PolyU Academy for Interdisciplinary Research (PAIR).

In addition to discussing the transformative potential of generative AI (GenAI) across various sectors, including healthcare, finance, and retail, highlighting its applications in areas such as medical imaging and fraud detection, Professor Yang also shared her personal AI milestones. They include Alibaba’s M6 model, which is capable of training a 10-trillion-parameter model using just 512 GPUs.

New model breaks new ground in AI development

One of the highlights of the keynote was an introduction to “InfiFusion” – her Co-GenAI project that creates a foundational model by combining smaller, stackable domain-specific models. Using this “model over models” approach, InfiFusion effectively consolidates knowledge from various sources, addressing vocabulary mismatches and enhancing computational efficiency.

In experiments, InfiFusion outperforms leading models, including Alibaba’s Qwen-2.5-14B-Instruct and Microsoft’s Phi-4, across multiple benchmark tasks. The novel training pipeline enhances AI accessibility while reducing reliance on centralised computing resources.

This groundbreaking research is a pivotal project for PolyU’s newly established AI+ Academy, which is led by Professor Yang. It will shift the focus of AI development from building extensive, general-purpose large language models (LLMs), to creating specialised models that can perform tasks with greater accuracy and efficiency. This pioneering initiative has placed Hong Kong and the Mainland at the leading edge of global progress in generative AI.

In line with these advancements, PolyU is dedicated to enhancing AI education and research. In January 2025, the University established the Faculty of Computer and Mathematical Sciences, aiming to spearhead global progress in digital transformation and AI through education, research, and knowledge transfer. This initiative further reinforces Hong Kong's leadership in the evolving AI technology landscape.

Prof. Chen Qingyan, Director of PAIR and Chair Professor of Building Thermal Science of the Department of Building Environment and Energy Engineering, gave a welcome speech at the Forum.

The panel discussion was led by Prof. Zhang Chenqi, Chair Professor of Artificial Intelligence of the Department of Data Science and Artificial Intelligence, and Director of the PolyU Shenzhen Research Institute (left), with Prof. Yang Hongxia (centre) and Prof. Li Qing, Head and Chair Professor of Data Science of the Department of Computing and Co-Director of the Research Centre for Digital Transformation of Tourism (right), offering their perspectives on the opportunities and challenges that AI advancements pose for higher education and research.