■ 演講時間: 114年10月14日(二) 10:00~11:30

■ 演講地點: E6-A203教室

■ 演講者: 郭宗杰講座教授 (南加大電機及計算機系)

■ 演講主題: Collaboration between Humans and Machines via Augmented Intelligence

■ 摘要：

The rapid advancement of AI, such as large language models (LLMs), has significantly impacted our society. Many questions arise. Will AI replace humans? Will machines control humans? How do we educate the next generation in the modern AI era? Confusions are partially caused by a lack of understanding of today’s AI tool - deep neural networks (DNNs). A DNN is a data-fitting machine that learns from many input/output pairs and predicts the output based on an input. Once we know the basic principle of DNNs, we can understand what they can and cannot do. I have researched an alternative solution to the data-fitting problem, called green learning (GL). It is “green” since it demands low power consumption in training and inference. Most importantly, GL helps demystify AI systems.

The human brain is a complex organ. Our understanding of the human brain is still limited. One difference between AI and human intelligence (HI) is that the latter can learn from very few examples. In machine learning terms, AI and HI are heavily and weakly supervised systems, respectively. The weakly supervised HI system allows humans to generalize well.

AI and HI can and will play a complementary role, leading to augmented intelligence. Humans need to leverage AI tools to increase productivity while avoiding their side effects (e.g., hallucination and emotional/intellectual dependence). Human brains will reach their full potential through education. This is the education mission in the modern AI era.