主任的話
Director’s Message
臺大人工智慧技術暨全幅健康照護聯合研究中心(亦稱：臺大人工智慧研究中心，AINTU)於2018年起，在科技部陳良基前部長推動「小國大戰略」的人工智慧國家政策下成立，希望整合國內大學和研究單位的AI學研能量，除致力AI研究外，並引領前瞻議題，推動國際交流，擴大成果效益，落實產業應用。臺大人工智慧研究中心是目前科技部AI專案中最大的研究中心，主題為人工智慧核心技術和醫療照護應用。
至今中心成立已滿三年，各項業務也上軌道。在109年度，轄下共有32個跨單位的AI計畫執行，約佔整體AI專案計畫43%。共有130餘名教授，740餘位研究人員參與計畫執行，是國內首屈一指的規模。中心今年與多個國際單位進行合作，包括與日本的AIST AIRC、RIKEN AIP、東京工業大學、史丹佛醫學院，並參與全球聯邦學習 EXAM計畫等。中心也持續培育人才，舉辦 AI Meetup、AI Summer School、Google AI Bootcamp、Medical AI School等活動；與中國信託，聿信醫療等多家廠商展開合作；並於11月舉辦成果展示，邀請科技部AI專案70多個團隊展出，也於12月的醫療科技展安排8個醫療照護團隊參展，推動研究成果與產業進行連結。此外中心也爭取到科技部拔尖整合計畫 (Capstone)，在臺大醫院急診部進行三年期的 “以人工智慧技術輔助醫療臨床決策與解決急診壅塞”AI模型開發，場域建置與落地試驗。
轄下團隊的研究成果也非常豐碩，包括廖弘源教授開發的最快最準的物件偵測模型YOLO-V4、徐宏民教授發表“xCos：可解釋AI人臉辨識技術”、賴飛羆教授發表“AECOPD發作預測系統”、黃升龍教授技轉的「高解析活體光學同調斷層掃描儀」獲得FDA醫材認證等，對學術界與產業界都有很大的影響。
展望未來，中心在2021年將主辦MLSS 2021，把國際優秀的AI人才吸引到臺灣來；也會加強與Google的深度研究合作，把臺灣的技術推廣到世界；在2023年，中心將主辦資料檢索領域頂級會議 ACM SIGIR，把該領域的世界專家帶到臺灣來。隨著世界各國在AI領域的進步，中心也會加入探討AI應用時面臨的議題，如個資隱私、資料自主、社會共好等，開發可信賴的AI技術與應用落地。臺大AI中心將持續扮演學界與業界、臺灣與國際、技術與應用的橋樑推手，讓AI成為臺灣的優勢，讓AI引發更多創意，讓AI能被信賴，讓AI將世界變得更好。
In 2018, under the MOST’s “Grand Strategy for a Small Country” national AI policy initiated by ex-Minister of Science and Technology Liang Gee Chen, the MOST Joint Research Center for AI Technology and All Vista Healthcare (also known as AI Research Center at NTU, AINTU) was founded. The AINTU’s goal is to become an enabler and to support cross-university research teams. With our support, each team can learn from others, exchange ideas with international partners, promote developed results, and implement them on the selected industrial use cases. AINTU is now the largest Research Center under MOST’s AI Innovation Program.
It’s been three years since the AINTU was established. In 2020, 32 cross-university projects are under execution, and the number of projects is around 43% of the AI Innovation Program. More than 130 professors (PI and co-PI) and 740 researchers were involved in these projects.
We also collaborated with several international institutions on organizing competitions, seminars, and researches, including with AIST AIRC, RIKEN AIP, Tokyo Institute of Technology, Stanford Medical School, and also attended the EMR Chest X-ray AI Model (EXAM) federated learning initiative.
At the same time, we keep cultivating talents by organizing academic events, such as AI Meetups, AI Summer School, Google AI Bootcamp, and Medical AI School. We also initiated several industry collaboration projects with partners, such as China Trust Commercial Bank, Heroic Faith Medical Science, etc. We also organized a large-scale technology demonstration in November, in which more than seventy project teams, sponsored by the MOST AI Innovation Program, were invited. In December, AINTU attended the annual Healthcare+ Expo together with eight project teams. In addition, our Capstone project proposal, “Transforming ED throughput with AI-Driven Clinical Decision Support System (CDSS): The Impact on the Delivery of Care and Patient Experience,” was accepted by MOST and start execution in March.
We also have fruitful research results from many projects, here are some examples: YOLO-V4, the fastest and most accurate object detection model developed by Professor Hong-Yuan Mark Liao, “xCos: an explainable face recognition technology” developed by Prof. Winston Hsu, “AECOPD: precise healthcare service for COPD patients“ developed by Prof. Feipei Lai, and the “ApolloVue S100: Optical Coherence Tomography Imaging System” received FDA 501(k) PN, in which the IP was transferred from Prof. Sheng-Lung Huang. We do believe these contributions brought significant impact to both academia and industries.
Looking to the future, we will attract AI talents to Taiwan by holding MLSS 2021 (Machine Learning Summer School) in 2021. We plan to promote our technologies to the world by leveraging stronger research collaborations between AINTU and Google. We successfully won the bid to host the top-tier conference ACM SIGIR 2023, and we expect to gather international professionals coming to Taiwan then. With the world AI trend, we want to cover more topics when putting AI technologies to real user scenarios, such as privacy, consumer data rights, AI for social good, trustworthy AI, etc. AINTU will play the bridging role between academia and industries, between Taiwan and foreign countries, between core technology and application scenario. Our goals are to make AI becoming the strength of Taiwan, to use AI to enable more innovations, to make AI trustworthy, and to use AI to create a better life.
