FSDC Career Day 2022 “Pitch Perfect” Video Competition

Background:
Financial Services Development Council (FSDC) Career Day 2022 – Pitch Perfect Video Competition aims to encourage the young people of Hong Kong to develop innovative ideas to promote Hong Kong as a Future Tech-lent Hub. The Topic for the competition was: “To promote Hong Kong as an international financial centre and develop the city into a global innovation and technology hub”.

Students were required to submit a two-to-five-minute video and an optional proposal. Shortlisted finalists were invited to attend the FSDC Career Day on September 24, 2022 (with their video played during the event).

Website / Image URL:
FSDC Career Day 2022 – Pitch Perfect Video Competition

Award(s):
Champion

Awardee(s):

Mr. Lau Pak Hei, BBA(Law)&LLB, Year 3

Students Sharing:
“I am honoured to be selected as the champion of the FSDC Career Day 2022 “Pitch Perfect” Video Competition. As a local student myself, the topic was particularly relevant to me as there is an observable trend of the exodus of professionals and our city slipping in competitiveness rankings, with cities like Dubai and Singapore quickly catching up. Therefore, I came up with a few strategies including promoting Hong Kong as an ESG investment hub, offshore RMB centre, innovation centre and most importantly reskilling and upskilling our local workforce. The Career Day was also a precious opportunity to mingle with panelists who were also from the crypto/web3 scene in Hong Kong.”
(by Lau Pak Hei)

Other Events
港大经管学院最新多模态AI图像生成能力排名出炉 部份中国人工智能模型表现突出
2025 | 学院新闻
港大经管学院最新多模态AI图像生成能力排名出炉 部份中国人工智能模型表现突出
港大经管学院今日发表《人工智能模型图像生成能力综合评测报告》,针对15个“文生图模型”及7个“多模态大语言模型”进行全面评估。研究显示,字节跳动的即梦AI和豆包,以及百度的文心一言,在新图像生成的内容质素及图像修改的表现突出;而早前引起全球关注的DeepSeek最新推出的文生图模型Janus-Pro,则在新图像生成方面表现欠佳。研究亦发现部分文生图模型虽然在内容质素方面表现优异,却在安全与责任方面的表现强差人意。整体而言,与文生图模型相比,多模态大语言模型整体表现较佳。
人工智能图像生成评测成绩单公布: 字节跳动百度表现亮眼,DeepSeek Janus-Pro表现欠佳
2025 | 研究
人工智能图像生成评测成绩单公布: 字节跳动百度表现亮眼,DeepSeek Janus-Pro表现欠佳
如今,人工智能领域的前沿模型技术已经从文本处理拓展至视觉信息的深度理解与生成。这些模型既能精准解读图像语义,又能根据文字描述创作出兼具真实感与艺术性的视觉内容,展现出令人惊叹的跨模态理解与创作能力。本研究聚焦全新图像的生成和基于现有图像的图像修改两大核心任务,提出了一套系统性的人工智能模型图像生成能力评测框架。我们基于多维测试集的构建与专家评审,对15个专业文生图模型和7个多模态大语言模型的图像生成能力进行了全面评估。结果显示,字节跳动的即梦AI和豆包以及百度的文心一言在新图像生成的内容质量与修改任务中表现突出,位列第一梯队。对比不同类型的AI模型,我们发现,相对于专业文生图模型,多模态大语言模型整体表现更佳。