人工智慧研究職業：從研究到現實 [S73849]

AI Research Careers: From Research to Reality [S73849]
Tomasz Bednarz,Director of Strategic Researcher Engagement, NVIDIA
Oya Celiktutan,Associate Professor, Reader in AI and Robotics, King's College London
Travis Waller,Professor, TU Dresden
Yannis Ioannidis,Professor, University of Athens
Peter Coveney,Professor, University College London

Hear a panel of pioneering AI researchers share their personal journeys and the transformative impact of their work. Get insights into how their research has revolutionized real-world applications of AI, from generative AI and robotics to digital twins to urban planning, and more. Discover the challenges and triumphs of AI research. Discover the pathways to success in AI research, garnering motivation and actionable advice from those who've shaped the field, ensuring that you'll leave inspired and equipped to embark on your own AI research journey.
Wednesday, Mar 19
5:00 PM - 6:10 PM CST

大家好，無論是早上、中午還是晚上，很高興歡迎各位參加我們的座談會，主題是「從研究到現實的人工智能研究旅程」(AI Research Careers: From Research to Reality)。我是托馬什·貝德納茨 (Tomasz Bednarz)，目前擔任輝達公司 (NVIDIA) 的策略研究參與主任 (Director of Strategic Researcher Engagement)。我很榮幸能主持這個由傑出研究者組成的座談會，他們正在多個領域推動人工智能 (AI) 的革新。我們將一起探討他們的開創性工作如何從生成式人工智能 (Generative AI)、機器人技術 (Robotics)、數位孿生 (Digital Twins)、城市規劃 (Urban Planning) 等領域，轉化為影響現實世界的應用。我們也會分享這些研究者的個人經歷、他們克服的挑戰，以及給予有志於投身人工智能研究的你們一些具體建議，讓你們帶著靈感與實用經驗離開。

Hello. Good afternoon, good morning, and evening, everyone, and welcome to our panel, "AI Research Careers: From Research to Reality." My name is Tomasz Bednarz, and I am currently the Director of Strategic Researcher Engagement at NVIDIA. I am delighted to moderate this distinguished panel featuring pioneering researchers who are advancing AI across multiple domains. We will explore how their groundbreaking work in fields such as generative AI, robotics, digital twins, and urban planning is transforming into real-world applications. We will also discuss their personal journeys, the challenges they’ve overcome, and practical advice for those of you aspiring to embark on a career in AI research, ensuring you leave inspired and equipped with actionable insights.

今天的對話將為我們帶來深刻的啟發。我們先從簡短的介紹開始。我想請教奧亞·切利克圖坦 (Oya Celiktutan) 副教授開場，她是倫敦國王學院 (King's College London) 的人工智能與機器人技術 (AI and Robotics) 領域的專家。請您開始吧。

The insights from our conversations today will be profoundly inspiring. Let’s start with a brief introduction. I’d like to begin with Associate Professor Oya Celiktutan, who is a reader in AI and robotics at King's College London. Over to you.

大家好，很高興能參加這場座談會。感謝主辦方的邀請。我是奧亞·切利克圖坦 (Oya Celiktutan)，在倫敦國王學院 (King's College London) 擔任人工智能與機器人技術 (AI and Robotics) 的副教授。我的職業生涯始於土耳其 (Turkey)，在那裡我完成了博士學位，研究領域是計算視覺 (Computer Vision)，特別專注於人類行為模式 (Human Behavior Modeling)。起初，我通過開發基於圖形的視覺模型 (Graph-based Vision Models) 來研究人類的面部表情、行為與情緒。後來，我的職業道路帶我前往法國 (France)，再到英國 (UK)。在英國，我將研究擴展到人類感知科學 (Human Perception Science)，並整合機器人技術與多模態數據 (Multimodal Data)。我的核心課題是探索如何將這些模型應用於機器人，使它們具備學習、行動與互動的能力。

Hi, everyone. It’s a pleasure to be on this panel. Thank you for the invitation. I am Oya Celiktutan, an associate professor in AI and robotics at King's College London. My career journey began in Turkey, where I completed my PhD in computer vision, focusing on human behavior modeling. Initially, I worked on understanding human facial expressions, behaviors, and emotions by developing graph-based vision models. My path then took me to France and later to the UK, where I expanded my research into human perception science, integrating it with robotics and multimodal data. My core focus is exploring how these models can be applied to robots, enabling them to learn, act, and interact.

在我的研究小組「社會人工智能實驗室」(Social AI Lab) 中，我們專注於機器學習 (Machine Learning) 與人機互動 (Human-Robot Interaction) 的交叉領域，並採取跨學科的方法來應對挑戰。我們致力於開發演算法 (Algorithms)，讓機器人能與人類及其環境無縫互動，支援日常生活中的各種需求。我們的研究挑戰了多重智能 (Multiple Intelligences) 的極限，結合視覺 (Visual)、運動 (Motor) 和人際 (Interpersonal) 能力。例如，我們探索機器人如何通過觀察他人來學習新任務，理解人類的非語言溝通技巧 (Non-verbal Communication Skills)，並提升它們在人類社會中的接受度。

In my research group, the Social AI Lab, we focus on the intersection of machine learning and human-robot interaction, tackling these challenges with an interdisciplinary approach. We work on developing algorithms that allow robots to seamlessly interact with humans and their environments, supporting various needs in daily life. Our research pushes the boundaries of multiple intelligences, combining visual, motor, and interpersonal capabilities. For instance, we investigate how robots can learn new tasks by observing others, comprehend human non-verbal communication skills, and increase their acceptance in human society.

我與產業及相關組織合作，利用互動式機器人 (Interactive Robots) 創造實際影響。例如，我們開發技術來協助行動不便的人士，幫助他們提升獨立性；或在醫療照護中支援兒童，減輕他們的焦慮並改善體驗。接下來，我想將發言權交給彼得·科文尼 (Peter Coveney)，他是倫敦大學學院 (University College London) 的教授，也是我的同事。請您接著分享。

I collaborate with industry and organizations to create real-world impact using interactive robots. For example, we develop technologies to assist individuals with limited mobility, helping them gain greater independence, or support children in healthcare settings to reduce anxiety and improve their experiences. I’d now like to hand over to Peter Coveney, a professor at University College London and my colleague. Over to you.

非常感謝。能參與這個座談會真是我的榮幸。我想先簡單分享一下我的背景，以及我是如何走到今天這一步的。我的經歷頗有意思。當我還是大學生並開始投入研究時，我深深著迷於物理科學 (Physical Sciences)、數學 (Mathematics) 和部分工程學 (Engineering)。那時我的研究焦點相當專精，甚至可以說有些冷門，主要屬於學術興趣。然而，在我獲得英國第一份長期職位後，我開始感到不安，因為我的興趣其實遠比這些單一領域來得廣泛。後來，我被說服加入一家當時非常有趣且知名的高科技服務公司，叫做桑伯格 (Schlumberger)。

Thank you very much. It’s an honor to join you all on this panel. I’d like to tell you a little bit about my background and how I came to do what I do today. It’s an interesting journey. When I was an undergraduate and first began doing research, I was heavily focused on the physical sciences, mathematics, and some elements of engineering. My work was quite specialized, even esoteric, and largely of academic interest. But after securing my first permanent appointment in the UK, I started to feel restless because my interests were broader than those narrow silos suggested. I was persuaded to join a very interesting and well-known company at the time, a high-tech service company called Schlumberger.

加入桑伯格 (Schlumberger) 時，這家公司以其高科技聲譽著稱，這一點深深吸引了我。我突然發現自己置身於一個全新的高科技環境，這是我之前從未接觸過的領域。在那裡，所有工作都以技術驅動。我最興奮的是在這樣的產業環境中，我們面對許多多元化的問題，而解決這些問題的方式是運用手邊一切可能的工具，這種方法類似工程思維 (Engineering Approach)。這意味著我們的工作本質上是跨學科的 (Interdisciplinary)，因為幾乎所有值得解決的問題都需要跨越多個領域的專業知識。此外，公司還有一股推動力，總是追求更大、更好、更快的目標，充分利用基礎設施和技術。

When I joined Schlumberger, it was distinguished in its domain by its high-tech reputation, which immediately drew me in. I found myself suddenly immersed in a kind of high-tech environment I had never engaged with before. Everything within that organization was driven by technology. What I found most exciting about working in such an industrial context was the diversity of problems we had to address. The approach was to solve those problems by any means at our disposal, which you might think of as an engineering approach. This meant the work was intrinsically interdisciplinary, as almost any problem worth solving required expertise across multiple domains. There was also a drive to always do things bigger, better, and faster by leveraging infrastructure and technology.

我在桑伯格待了大約八年。那是 1990 年代，你們可能記得，那是平行計算 (Parallel Computing) 和高效能計算 (High Performance Computing) 開始興起的時代。桑伯格成為首個非國防相關組織購買高效能電腦的公司，那是一台來自思考機器公司 (Thinking Machines Corporation) 的連接機器 (Connection Machine)。這台設備讓我們得以在多個領域進行有趣的探索，包括如今被稱為人工智能 (Artificial Intelligence) 的技術。當時正值所謂的「AI 寒冬」(AI Winter)，因為幾年前的誇大宣傳導致期望落空，但我們其實正在為類似神經網絡 (Neural Networks) 的技術奠定基礎。我們將這些技術廣泛應用於各種問題。

I spent about eight years at Schlumberger. This was during the 1990s, an era you may recall as the time when parallel computing and high performance computing began to take off. The company became the first non-defense-related organization to purchase a high-performance computer, a Connection Machine from Thinking Machines Corporation. That acquisition led us to do many interesting things across diverse areas, including what is now called artificial intelligence. At the time, we were living through one of those AI winters caused by exaggerated claims from earlier years, but we were essentially laying the foundations for technologies like neural networks. We applied them widely across various challenges.

當時我堅信這是公司應該走的路，但在當時這些技術並未獲得如今這樣的關注。然而，這段經歷深刻影響了我。後來，我回到學術界，如簡介中提到的，自那時起我一直在倫敦大學學院 (University College London) 領導計算科學中心 (Centre for Computational Science)。這個中心延續了我一貫的研究方式：跨學科、技術驅動，並致力於解決現實世界的問題。如今我們結合超高效能超級計算 (Ultra High-end Supercomputing)、人工智能 (AI) 和量子計算 (Quantum Computing)，應用於從物理學到醫學的領域。我很高興能將產業經驗帶入學術研究，並持續推動科學進步。接下來，我將把發言權交給下一位講者。

Back then, I was convinced this was the direction the company should pursue, but interestingly, it didn’t gain the traction it has since achieved. That experience profoundly shaped me. Later, I returned to academia, and as described in the slides, I’ve been running what I call the Centre for Computational Science at University College London ever since. This center embodies the same approach: interdisciplinary, technology-driven, and focused on solving real-world problems. Today, we combine ultra high-end supercomputing, artificial intelligence, and quantum computing, applying them to fields ranging from physics to medicine. I’m thrilled to bring my industry experience into academic research and continue pushing scientific progress. Next, I’ll hand over to the following speaker.

我們致力於解決各式各樣的問題。我們的口號是「通過電腦推進科學」(Advancing Science Through Computers)。這個口號並未限定我們的研究領域。在今天的討論中，你們可能已經注意到，我們的工作很大一部分——但並非全部——集中在計算生物醫學 (Computational Biomedicine)。隨著數位化 (Digitalization)、大規模數據集 (Large Datasets) 和超級計算 (Supercomputing) 的發展，這個領域正變得越來越引人注目且可行。這些技術的進步，例如如今價格低廉的 GPU (Graphics Processing Units)，在過去是無法想像的。正是這些背景推動我走到今天的位置。

We look to solve problems of any sort. Our catchphrase is "Advancing Science Through Computers." It doesn’t specify which areas we work on. In the context of this discussion, it may be clear that quite a lot of what we do—though by no means all of it—is in the area we call computational biomedicine. This field is becoming increasingly compelling and tractable with the advent of digitalization, large datasets, and the ability to exploit supercomputing. Technologies like affordable GPUs, which weren’t around long ago, have played a key role. It’s the context of all these developments that has brought me to where I am today.

我現在與來自不同領域的人密切合作。我的博士生來自各種具有量化背景 (Quantitative Background) 的學科。我們在國際範圍內協作，解決有趣的問題。我們在某些領域學到的知識可以迅速轉移到其他領域，因為這些演算法 (Algorithms) 有許多共通點。對我來說，保持開放的態度並善於合作，是我們能做到今天這些事情的關鍵。這也反映了我們的研究方式。我的國際合作始於我在產業界的日子，那時我在桑伯格 (Schlumberger) 這樣一家高度國際化的公司工作。如今，我與美國的同事保持緊密聯繫，利用全球頂尖的超級電腦，例如美國能源部 (DOE) 的設施，來推動我們的計畫。

I now work closely with people across many different domains. My PhD students come from any area with a quantitative background, and we collaborate on an international scale to solve interesting problems. What we learn in some areas becomes rapidly transferable to others because the algorithms share many common features. For me, being open to opportunities and having the ability to collaborate have led us to do what we do today. This approach is illustrative of our work. My international activities date back to my time in industry at Schlumberger, a highly international company. That continues today with strong ties to colleagues in the USA, where we access some of the world’s highest-performance supercomputers through the Department of Energy (DOE).

其中一個例子是我們開發的程式碼 HemeLB，這是針對血流 (Blood Flow) 的模擬工具。我們現在可以模擬整個人體的血管系統 (Vascular System)，從頭到腳，包括動脈和靜脈。我們能在單一案例中使用像 Frontier 這樣的超級電腦，這台機器擁有大約一萬個節點 (Nodes)，每個節點配備八個 GPU，總計超過八萬個 GPU 進行運算。這樣的規模讓模擬速度極快。我們的目標不僅止於此，我們希望將各個器官模型 (Organ Models) 連接起來。例如，我們已經能將心臟模型 (Heart Model) 與血管系統相連，模擬心血管系統 (Cardiovascular System)。這開啟了許多當前醫學尚未深入探討的問題，例如血液、心臟和大腦之間的相互作用。

One example is a code we work on called HemeLB, designed for blood flow simulation. We can now model and simulate the entire human vascular system from head to toe, including arteries and veins. For a single instance, we might use the entirety of Frontier, a machine with roughly 10,000 nodes, each equipped with eight GPUs per node, allowing us to run applications across over 80,000 GPUs. This scale makes it incredibly fast. In this context, we are ambitious about connecting organ models to each other. For example, we can connect a heart model to the vasculature. Before you know it, we’re simulating the cardiovascular system, opening up interesting questions often not even considered in current medicine, such as the coupling between processes in the blood, the heart, and the brain.

這項研究的潛力在於將計算科學應用於實際醫學挑戰。我希望這些努力能為健康科學帶來突破。說完了我的部分，我想將發言權交給下一位講者，史蒂芬·特拉維斯·沃勒 (S. Travis Waller)。他是德國德累斯頓工業大學 (Technical University of Dresden) 的教授。請您接著分享。

The potential of this work lies in applying computational science to real medical challenges. I hope these efforts will lead to breakthroughs in health sciences. Having said my piece, I’d like to pass the floor to Stephen Travis Waller, a professor at the Technical University of Dresden in Germany. Over to you.

謝謝你，彼得。我是特拉維斯·沃勒 (S. Travis Waller)，在德國德累斯頓工業大學 (Technical University of Dresden) 擔任交通建模與模擬講座教授 (Chair of Transport Modelling and Simulation)，同時也是燈塔教授 (Lighthouse Professor)。我搬到這裡已經大約三年了。在此之前，我在澳洲雪梨新南威爾斯大學 (University of New South Wales, UNSW) 擔任土木與環境工程學院 (School of Civil and Environmental Engineering) 的教授兼院長，從 2011 年到我搬遷前都在那裡。更早之前，我在德克薩斯大學奧斯汀分校 (University of Texas at Austin) 任教約九年，還曾在伊利諾大學香檳分校 (University of Illinois Urbana-Champaign) 和西北大學 (Northwestern University) 做博士後研究。我的學術背景始於電機工程 (Electrical Engineering)，後來在博士研究中轉向工業工程與管理科學 (Industrial Engineering and Management Science)，專注於複雜系統的建模 (Modeling Complex Systems)，特別是社會技術系統 (Socio-technical Systems)。

Thank you, Peter. My name is Travis Waller. I’m a Lighthouse Professor and Chair of Transport Modelling and Simulation at the Technical University of Dresden in Germany. I’ve lived here for about three years now. Prior to that, I was a professor and Head of the School of Civil and Environmental Engineering at the University of New South Wales in Sydney, Australia, from around 2011 until my relocation. Before that, I was on faculty at the University of Texas at Austin for about nine years. I’ve also been at Urbana-Champaign and did postdoctoral work at Northwestern. My background started in electrical engineering, and then for my doctoral work, I moved into industrial engineering and management science. So, it’s very much about modeling complex systems, typically socio-technical systems.

過去二十年，我的興趣逐漸轉向圖論 (Graph Theory)、優化 (Optimization)、機器學習 (Machine Learning) 和數據科學 (Data Science)。我發現自己有兩個核心關注點：一是新興技術 (Emerging Technologies) 如何影響人類移動 (Human Mobility)，包括改變旅行行為 (Travel Behavior) 本身，以及如何利用這些技術更好地表達複雜系統；二是社會意識的演進 (Evolving Social Consciousness)，例如環境影響 (Environmental Impact)、正義 (Justice)、公平 (Equity) 和公正 (Fairness)，並探索如何將這些因素整合到模型中，以提升我們的表現。我的研究既有高度理論性，像許多同事一樣，同時也非常實用。我們獲得了各國國家科學基金會 (National Science Foundations) 的資助，也與許多實務機構和應用單位合作，將研究成果付諸實踐。

Over the last 20 years, my work has increasingly branched into graph theory, optimization, machine learning, and data science. I’ve discovered two key interests: first, how emerging technologies impact human mobility, both in terms of changing travel behavior itself and leveraging those technologies to better represent complex systems; and second, our evolving social consciousness—things like environmental impact, justice, equity, and fairness—and how we can integrate these into models to do a better job. My work is highly theoretical, like that of many of my colleagues, but also very practical. We’ve been funded by national science foundations across countries, as well as by numerous practicing agencies and institutions, deploying our work in real-world settings.

你們可以看到我過去帶領的團隊，包括德克薩斯大學奧斯汀分校、新南威爾斯大學，以及現在的德國團隊。與輝達 (NVIDIA) 的合作已經持續多年。我們開發了人工智能驅動的建模技術 (AI-driven Modeling)，試圖超越歷史上的限制，快速構建這些模型。過去，這些模型需要大量時間和資源，但新興數據 (Emerging Data) 的爆炸性增長改變了一切。這兩者的結合帶來了轉型，讓我們能做到過去無法想像的事情。例如，我們研究了人類衝突期間的旅行行為 (Travel Behavior During Human Conflict)、跨國建模 (Multinational Modeling)，這在過去非常困難；我們也關注災後需求評估 (Post-disaster Needs Assessment)，最近的案例是 2024 年羅馬尼亞洪水 (2024 Romanian Floods)。

You can see some of the teams I’ve led over the years at the University of Texas at Austin, UNSW, and now here in Germany. Our collaboration with NVIDIA has been ongoing for years. We’ve been developing AI-driven modeling to transcend what’s been possible historically, rapidly building these models that used to take a great deal of time and resources. The tremendous opening up of emerging data has been transformative. These two things in combination have allowed a transformation in what we can do. For instance, we’ve looked at travel behavior during human conflict and multinational modeling, which historically was very problematic. We’ve also worked on post-disaster needs assessment, most recently with the 2024 Romanian floods.

我們試圖將這些研究整合到輝達平台 (NVIDIA Platforms) 的強大功能中。我們專注於系統的表達 (Representation of the System)，而不是視覺化 (Visualization) 或公眾參與 (Engagement) 的專家，但這些對與公眾和決策者溝通至關重要。因此，我們努力將這些不同元素整合起來，既確保模型在科學上的代表性，又讓它易於理解，以支持決策並改變未來。這就是我們的概況。接下來，我想將發言權交給我的同事揚尼斯·約安尼迪斯 (Yannis Ioannidis)，他是雅典大學 (University of Athens) 的教授。請您分享。

We’ve been trying to integrate all of this into the vast capabilities of NVIDIA platforms. We focus heavily on the representation of the system. We’re not specialists in visualization or engagement, but these are critical for connecting with the public and decision-makers. So, we’re working hard to align all these different components to make our models both scientifically representative and accessible, helping inform decisions and shape the future. That’s us in a nutshell. With that, I’d like to pass over to my colleague Yannis Ioannidis, a professor at the University of Athens. Over to you.

非常感謝。能參加這個座談會我感到非常榮幸。我要再次感謝主辦方的邀請。我想簡單介紹一下我的背景。我的本科學位是電機工程 (Electrical Engineering)，之後我前往美國攻讀研究生課程。我從應用數學 (Applied Mathematics) 開始，後來轉向電腦科學 (Computer Science)。我在加州大學柏克萊分校 (UC Berkeley) 和威斯康辛大學麥迪遜分校 (University of Wisconsin, Madison) 完成學業，隨後在威斯康辛大學電腦科學系任教 11 年，之後回到希臘。我在希臘已經待了將近 30 年。

Thank you very much. It’s a great pleasure to be part of this panel. Let me thank the organizers again for this invitation. I’ll briefly share my background. I won’t go into too much detail, except to say that my undergraduate degree was in electrical engineering. Then I went to the US for graduate studies, starting with applied mathematics and eventually moving into computer science. I studied at Berkeley and then the University of Wisconsin, Madison, where I stayed as a faculty member in computer science for 11 years before returning to Greece. I’ve been here for almost 30 years now.

接下來，我想介紹我目前扮演的幾個角色，這些角色試圖涵蓋學術生涯中可能涉及的各個面向。我仍在不斷探索和完善這些角色。第一個角色，也是其他角色的頂點，是我擔任計算機協會 (Association for Computing Machinery, ACM) 的主席，目前已連任第二屆。ACM 是全球最大的計算專業人士組織，擁有約 12 萬名成員。這是一項極具挑戰性的任務，不僅因為我們領域的豐富性，還因為計算技術和人工智能 (Artificial Intelligence) 在當今扮演的角色帶來了諸多挑戰。制定 ACM 的未來策略，幫助我們作為計算和 AI 專業人士應對變化，需要我從其他職位中累積的所有經驗與智慧。我希望自己能勝任這份工作。

Next, I’d like to describe the hats I’m wearing now, which try to capture all the aspects an academic might be involved in. I still have more to cover. The first one, if you will, is the culmination of all the others that have brought me here: being the President of ACM, now elected for a second term. ACM is the largest global organization of computing professionals, with about 120,000 colleagues. It’s an extremely challenging task, not only because of the richness of our field but also due to the challenges arising from the role computing and artificial intelligence play right now. Setting a strategy for how ACM can evolve and support us all as computing and AI professionals requires all the experience and wisdom that my other positions bring me. Hopefully, I’m doing a decent job at that.

雖然擔任 ACM 主席是一項志願工作，但它給了我全球影響力的舞台，也是我目前最興奮的部分。而我的正式工作是在雅典大學 (University of Athens)。從研究生階段開始，我的領域一直是數據管理 (Data Management)。我在雅典大學帶領「數據、資訊與知識管理小組」(Management of Data, Information, and Knowledge Group)，縮寫是 MaDgIK，我們希望能在數據管理領域帶來一些「魔法」(Magic)。這個職位，以及我在威斯康辛大學的經歷，教會我如何教學、研究並建立大型團隊。後來，我進入行政領域，在雅典娜研究中心 (Athena Research Center) 擔任了 10 年的領導角色，至今仍與其保持聯繫。

While being ACM President is a volunteer role, it offers a global footprint and is the most exciting part of what I do. My paying job, however, is at the University of Athens. Since the beginning of my graduate studies, my area has more or less been data management. I lead the Management of Data, Information, and Knowledge Group at the university—you can pronounce the acronym MaDgIK. Hopefully, we bring some magic into the field of data management. That position, along with my background at Wisconsin, taught me how to teach, conduct research, and build large teams. Later, I moved into administration, serving for 10 years at the Athena Research Center in Greece, where I’m still affiliated.

我在雅典娜研究中心 (Athena Research Center) 擔任了 10 年的主席。在那裡，我發現了一些非技術性的挑戰，這些挑戰是你需要面對的，以便凝聚研究人員、服務數據研究者和產業研究者，滿足整個機構的需求。對於許多大型科學領域，無論是否與計算科學 (Computing Science) 相關，都需要強大的基礎設施。因此，開放性 (Openness) 和開放存取 (Open Access) 成為關鍵。OpenAIRE 數據基礎設施 (OpenAIRE Data Infrastructure) 為歐洲帶來了這樣的改變，讓歐洲研究人員產出的所有研究成果、論文等得以公開。我帶領 OpenAIRE 超過 10 年，這段時間裡，我關注的不是自己的研究，而是如何通過基礎設施服務支持他人的研究。這是我從其他職位中學到的經驗教訓。

For 10 years, I was the president of the Athena Research Center. There, I discovered the non-technical aspects you need to bring together and serve researchers—data researchers, industrial researchers, and the entire institution. For many large sciences, whether computing science or not, you need large infrastructure. So openness and open access were key. The OpenAIRE infrastructure brought this to Europe for all the research results, papers, and more produced by Europeans. I led OpenAIRE for more than 10 years. During that time, it wasn’t about my own research but about serving the research of others through infrastructure—an approach that complemented what I’d learned from my other positions.

我的跨學科經驗在下一頁簡報中會更明顯。我參與了聯合國永續發展解決方案網絡 (UN Sustainable Development Solutions Network) 的全球氣候中心 (Global Climate Hub)，擔任聯合主席。這結合了跨學科 (Interdisciplinary) 的工作。此外，我還涉足產業，參與了幾家新創公司，其中一家是基於人工智能 (AI) 和數據驅動 (Data-driven) 的平台。這些經驗匯聚在一起，幫助我以 ACM 主席的身份成為整個社群的啟發性領袖。我希望這些多元角色能讓我更好地服務全球計算專業人士。

My interdisciplinary experience, which I’ll emphasize in the next slide, is evident in my involvement with the Sustainable Development Goals. I co-chair the Global Climate Hub of the United Nations Sustainable Development Solutions Network, which combines interdisciplinary work. Beyond that, I’m involved in industry through a couple of startups, one of which is an AI-driven, data-driven platform. All these experiences come together, helping me serve as an inspiring leader for the community as the President of ACM. I hope these diverse roles enable me to better support computing professionals globally.

我的工作核心主要圍繞數據基礎設施 (Data Infrastructures)、數據科學 (Data Science) 和數據與文本分析 (Data and Text Analytics)。過去 10 年左右，我開始被數據的人文面向吸引。數據本身可能很枯燥，但推薦系統 (Recommender Systems) 和個人化 (Personalization) 卻是關鍵領域，這些既有 AI 驅動的部分，也有非 AI 的方法。這也成為我進入互動式數位敘事 (Interactive Digital Storytelling) 的途徑。如何通過知識與人類互動、激發人類參與，是一件非常令人興奮的事情。跨學科幾乎從我成為教職員開始就一直是我的特點。我的研究，無論是數據管理還是數位敘事，都受到生命科學 (Life Sciences)、物理科學 (Physical Sciences)、社會科學 (Social Sciences)、人文學科 (Humanities) 和藝術 (Arts)——如舞蹈和繪畫——的啟發。與這些領域的同事互動讓我感到充實。

My work profile, as I mentioned, primarily focuses on data infrastructures, data science, and data and text analytics. Over the past 10 years or so, I’ve been drawn to the human side of data. Data can be dry, but recommender systems and personalization are key areas, driven by both AI and non-AI approaches. This became a pathway into interactive digital storytelling. How you engage humans with knowledge is something truly exciting. Interdisciplinarity has been part of my work almost from the beginning as a faculty member. My work, whether in data or storytelling, is inspired by many other fields—life sciences, physical sciences, social sciences, the humanities, and the arts, like dance and painting. I feel enriched by all the colleagues I interact with across these domains.

我想分享幾個目前的研究亮點。其中一個是，如何將程式語言與存在超過 30 或 40 年的主流資料庫語言 SQL (Structured Query Language) 結合起來。這就像油和水一樣，難以融合，但我們試圖找到一個有凝聚力的解決方案。這項工作展示了如何將傳統技術與現代需求整合。我的研究旅程一直是跨領域的旅程，而這些計畫讓我能夠將數據科學的技術力量與人類的需求和創造力連結起來。

Some highlights of my current projects: one is about combining programming languages with the main database language that’s been around for over 30 or 40 years—SQL. It’s like oil and water. How do you bring them together in a cohesive way? This work demonstrates how we can integrate traditional technologies with modern needs. My research journey has always been interdisciplinary, and these projects allow me to connect the technical power of data science with human needs and creativity.

講者4（揚尼斯·約安尼迪斯）的演說後續
在醫療數據 (Medical Data) 的聯合學習 (Federated Learning) 中，當隱私成為一大考量時尤為重要，因為這些是涉及人類的數據，必須尊重患者的隱私權。我提到的講故事體驗 (Storytelling Experiences)，主要是改變參觀博物館的整體哲學，但也適用於其他需要人類參與的環境。我們追求的是參與工程 (Engagement Engineering)，探索如何讓人們在情感上融入知識。而 OpenAIRE，如我之前所述，試圖捕捉整個研究成果，不僅限於歐洲研究人員，已涵蓋超過 2 億份出版物 (Publications)、7500 萬個數據集 (Datasets)。我們希望將這些整合成一個知識庫 (Knowledge Base)，讓其他研究者能夠加以利用。

Federated learning with medical data is crucial when privacy is a concern, because these are human data and you have to respect patients’ privacy. The storytelling experiences I mentioned are primarily about changing the overall philosophy of how you visit a museum, but they also apply to many other environments where humans need to be engaged. We aim for engagement engineering—how do you make people emotionally involved with knowledge? As for OpenAIRE, as I said, it tries to capture the whole production, not just of European researchers—more than 200 million publications, 75 million datasets, and so on. We work to bring them into a knowledge base so others can conduct research with them.

這就是我的個人檔案 (Profile)、我的道路，以及我的熱情所在。這些經歷和興趣最終帶我來到這裡，參與這場座談會。

This is, if you want, my profile—a personal path and my passions. That’s what brought me here.

講者1（托馬什·貝德納茨）與講者4的對話
托馬什·貝德納茨：

謝謝你，揚尼斯。我認為你今天的分享非常精彩，讓我們有機會跟你進行一些深入對話。你是 ACM 的主席，對吧？多年來，你親眼見證了計算研究的革命。能否多告訴我們一些，從你描述的資料庫系統 (Database Systems) 到領導 ACM，你的旅程如何塑造了你對人工智能研究職業 (AI Research Careers) 的看法？

Tomasz Bednarz:

Yes, thank you, Yannis. I think what you’ve shared here makes for some fascinating conversations with you right now. You’re the ACM President, right? So you’ve seen the revolutions of computing research firsthand over the years. Can you tell us a bit more? How has your journey from database systems, as you described, to leading ACM shaped your perspective on AI research careers?

揚尼斯·約安尼迪斯：

當我開始在資料庫 (Databases) 領域的研究生涯時，人工智能 (AI) 主要是自上而下 (Top-down) 的方法，依賴規則 (Rules)、框架 (Frames) 和基於物理學或其他選項的手動創建模型 (Manually Created Models)。這種方法取得了一些局部成功，自下而上 (Bottom-up) 的方法也存在，但因為計算能力 (Computational Power) 和數據規模 (Data Size) 的限制，成果非常有限。那時，AI 被視為某種異國情調的旁支，大多數人瞧不起它，覺得從事這領域的人有些奇怪。然而，隨著數據和計算能力與速度的增長，一切都改變了。如今，每個人都專注於自下而上的 AI，甚至圖靈獎 (Turing Awards) 在幾十年後又開始頒給 AI 的成就。

Yannis Ioannidis:

When I started my research career in databases, AI was mostly top-down—rules, frames, and manually created models based on physics or other approaches. It had some partial success, and bottom-up AI did exist, but it couldn’t deliver much because computational power and data size were limited. Back then, AI results were very restricted, and most people looked down on it. It was something more exotic on the side, and those working on it were seen as a bit odd. But as data, computational capacity, and speed grew, things changed. Now everyone is focused on bottom-up AI, and even Turing Awards are being given again for AI achievements after decades.

現在是 AI 的輝煌時代。之前有人提到「AI 寒冬」(AI Winter)，但現在是盛夏 (Summer)，而且這夏天可能會持續下去。不管人們從事什麼研究，他們很可能會說：「好吧，我也在做 AI。」這是為了不被時代拋下。我的觀點也隨之改變，我相信這種失衡最終會恢復。自下而上 AI 的弱點會逐漸顯現，我們正在努力找出這些弱點。雖然我主要專注於數據 (Data)，但我經常被邀請談論 AI 的各種話題。誠然，我的領域是數據，但現代 AI 完全依賴數據，所以我也能正當地說自己參與其中。這條路改變了我的視野，現在 AI 讓我非常興奮，它由數據驅動，充滿無限可能。

Now it’s the glorious time for AI. The winter was mentioned earlier, but now it’s the heart of summer, and it will probably continue. No matter what anyone is doing, they’re likely to say, “Okay, I’m in AI,” just to not feel left out. My perspective has also changed, and I believe balance will be restored. The weaknesses of this bottom-up approach to AI will show, and we’re trying to find them. Although I’m primarily a data person, I’m constantly invited to talk about AI this and AI that. My area, as I said, is data, but let’s face it—since modern AI depends completely on data, I’m okay saying I’m in AI too. This path has changed me. Now AI is extremely exciting, fed by data, and incredibly inspiring.

講者1（托馬什·貝德納茨）
非常感謝。正如俗話所說，一切都與其他事物相連，對吧？我記得多年前達文西 (Da Vinci) 就提出過這樣的觀點。好，讓我們繼續。正如我們所知，你的工作高度聚焦於社交機器人學 (Social Robotics) 和以人為中心的人工智能 (Human-centered AI)，這些都是令人興奮的主題。是什麼啟發你追求 AI 與機器人技術的特定交叉領域？能否多跟我們分享一些？還有，你的國際學術背景如何影響你的研究方法？

Thank you so much. It’s like they say—everything connects to everything else, right? I think it was a thought from Da Vinci many years ago. Alright, let’s continue. As we already know, your work focuses heavily on social robotics and human-centered AI, exciting topics. What inspired you to pursue this specific intersection of AI and robotics? Can you tell us a bit more about that? And how has your international academic background influenced your approach?

講者2（奧亞·切利克圖坦）
感謝你的問題。事實上，我非常認同這個問題，因為啟發我在這個領域工作的最大動力正是它的多學科性 (Multidisciplinarity)。人類與機器人的互動 (Human-Robot Interaction) 極其複雜。對我來說，人類行為 (Human Behavior) 是廣泛、多樣且常常難以預測的。要應對這些挑戰，我們必須整合多個學科，不僅是機器人技術 (Robotics) 和人工智能 (AI)，還包括社會科學 (Social Sciences)，例如行為心理學 (Behavioral Psychology) 和倫理學 (Ethics)。這一點至關重要，因為我們開發的方法需要保護隱私 (Privacy-preserving) 並符合倫理標準 (Ethically Sound)。

Thanks for the question. Actually, I couldn’t agree more with asking this, because what inspires me most about working in this domain is its multidisciplinarity. Human-robot interaction is extremely complex. For me, human behavior is vast, diverse, and often unpredictable. To tackle these challenges, we must integrate multiple disciplines—not just robotics and AI, but also social sciences like behavioral psychology and ethics. This is crucial, as the approaches we develop need to be privacy-preserving and ethically sound.

在不同的學術環境中工作，為我提供了與來自多元學科的研究者合作的寶貴機會，包括社會科學家 (Social Scientists)、臨床醫生 (Clinicians)，甚至藝術家 (Artists)。從這些不同視角學習，並將新想法應用到我的研究中，是我在這個領域保持熱情的動力。我非常享受這種感覺，將藝術 (Art) 與科學 (Science) 結合，推動機器人技術和以人為中心的 AI 發展，真是太棒了！

Working in different academic environments has offered me exciting opportunities to collaborate with researchers from diverse disciplines, including social scientists, clinicians, and even artists. Learning from these perspectives and applying new ideas to my work is what keeps me motivated in this field. It feels great—I love it. Bringing art and science together to advance robotics and human-centered AI is fantastic!

講者1（托馬什·貝德納茨）
非常感謝你，奧亞。接下來，特拉維斯 (Travis)，你的研究獲得了來自世界各地組織的大量資金支持，像是國家科學基金會 (National Science Foundations)。你一直在全球範圍內工作，對吧？

Thank you so much, Oya. Now, Travis, you’ve secured significant funding from organizations worldwide, like national science foundations. You’ve been working across the globe, right?

講者1（托馬什·貝德納茨）
特拉維斯，能否與我們分享你研究旅程中的一個關鍵時刻，讓你確信人工智能 (AI) 對交通系統 (Transportation Systems) 的變革潛力？

Could you share with us a pivotal moment in your research journey that convinced you of AI’s transformative potential for transportation systems?

講者3（特拉維斯·沃勒）
當然可以。這一切其實始於我的研究生階段。我的研究生學習非常專注於交通流動系統的建模 (Modeling of Transport Mobility Systems)，目的是為城市、區域和國家層面的規劃提供資訊。我們開發了一些新技術，甚至與我的博士導師從西北大學 (Northwestern University) 分拆出來，創辦了一家新創公司。這家公司運營了大約 15 年，還實現了盈利，表現得相當不錯。但它沒有進一步擴大的原因在於，當時創建這些模型非常客製化 (Bespoke)，極其依賴時間和資源，還需要大量的人為決策 (Human Decision Making)。這限制了模型的適用性和推廣。

Absolutely. It really began in my own graduate studies. My graduate work was very focused on modeling transport mobility systems to inform planning at city, regional, and national levels. We developed some new techniques and even spun those off out of Northwestern University with my doctoral supervisor. We created a startup, and it ran for about 15 years, made a profit, and did well enough. But the reason it didn’t grow bigger was that it was still very limited—creating those models was bespoke, extremely time- and resource-dependent, and relied heavily on human decision-making, which constrained its applicability and uptake.

後來，我的團隊開始研究新興數據集 (Emerging Datasets) 和應用人工智能 (Applied AI)。我們開發出更新的方法，將原本需要一年的建模時間縮短到幾週、幾天，甚至幾小時。這徹底改變了應用領域。我之前提到過在烏克蘭戰爭期間的旅行行為建模 (Modeling Travel Behavior During the Ukraine War) 和災後需求評估 (Post-disaster Needs Assessments)，這些應用在舊方法下是不可能的。但現在，這些新技術開啟了全新的可能性。這是真正的變革 (Transformative)。

Subsequently, as my research team evolved, we began learning more about emerging datasets and applied AI. We came up with newer methods that, instead of taking a year, can now be done in potentially weeks, days, or sometimes even hours. That’s radically changed the application space. I mentioned modeling travel behavior during the Ukraine war and post-disaster needs assessments—these sorts of applications were impossible with the old methods. But all of that opens up now with these new techniques. It’s transformative.

講者1（托馬什·貝德納茨）
太棒了！這些最新的計算平台真是令人驚嘆，對吧？我們在 GTC (NVIDIA GTC) 期間將會遇到許多挑戰，我相信大家都對此感到非常興奮。接下來，讓我們轉向彼得 (Peter)。彼得，我很欣賞你的工作。你也採用了跨學科 (Interdisciplinary) 的方法。正如我們之前描述的，你的計算生物醫學 (Computational Biomedicine) 研究從物理學 (Physics) 跨越到醫學 (Medicine)。

That’s awesome. And these newest computational platforms are amazing, right? We’re going to be tackling lots of problems during GTC, and I think everybody is super excited about that. Now, let’s move to Peter. Peter, I love your work. You have a disciplinary approach as well. Your computational biomedicine work spans from physics to medicine, as we described before.

你是如何在這些不同學科和領域之間進行轉換的？對於希望跨領域研究的學者，你有什麼建議？

How did you navigate the transitions between those different disciplines and domains altogether? And what advice would you give to researchers looking to work across these fields?

講者2（彼得·科文尼）
要在這些領域取得實質性進展，最大的挑戰在於真正讓事情發生。如果你從事計算生物醫學 (Computational Biomedicine)，我們的終極目標是在現代背景下使用數位孿生 (Digital Twins)。我們試圖創造個人的高保真表達 (High Fidelity Representations)。這結合了大規模計算 (Large-scale Computing)、我們能做的加速技術，以及人工智能 (AI) 的重要角色。這是一個涉及整個醫療社群的努力，因為它影響到我們每個人，尤其是醫療領域的專業人士和臨床醫生 (Clinicians)。然而，這些人通常未受過訓練去理解這些方法。從物理學 (Physics) 到化學 (Chemistry)、材料科學 (Materials Science)，再到生命科學 (Life Sciences) 的交界處，科學領域存在某種分隔。有一群人非常熟悉預測 (Prediction) 的作用，但這在醫療領域並不常見。

These are the biggest challenges to actually get substantial things happening in those areas. If you try to do computational biomedicine, ultimately we are aspiring in the modern context to use what we call digital twins. We’re trying to produce high-fidelity representations of individuals. That’s an interesting combination of large-scale computing, what we can do to accelerate it, and AI, which plays a substantial role there, with a community that involves everyone at the healthcare level because it affects us all. But in particular, a group of professionals in the medical domain and clinicians aren’t really trained to understand these methods. There’s a division across the sciences as we move from physics through chemistry and materials to the interface with the life sciences. And where is that group of people who are very familiar with the role of prediction?

預測意味著什麼？我對我研究的系統有足夠的了解，能夠預測可能發生的事情。如果我使用適當的方法——我們稱之為驗證 (Validation)、確認 (Verification) 和不確定性量化 (Uncertainty Quantification)——我就能在一定程度上證明這些預測是可靠的。這樣人們才會重視它們，就像天氣預報 (Weather Forecasting) 和某種程度上的氣候預測 (Climate Prediction) 一樣，都是這類方法的一部分。我們與理解這些方法論的人合作。然而，當你進入生命科學和醫學領域時，這些人並未被教育或培養出認為存在預測性模型 (Predictive Models) 的思維。但如果我們要實現個人化醫療 (Personalized Medicine)，這正是我們必須做到的——提前提供可供臨床醫生採取行動的資訊。這是一個相當大的挑戰，需要讓這些人接受我們的觀點。你不能低估教育和培訓的重要性，我們必須投入這些努力，讓他們以類似的方式看待世界，進而擴大這些方法的吸引力。

What does prediction mean? I know enough and understand enough about the system I’m studying to forecast something that might happen. If I use the appropriate methods, which we call validation, verification, and uncertainty quantification, I can get those predictions certified as reliable to some degree. Then people pay attention to them because weather forecasting, and to a lesser degree climate prediction, are all part of that. We’re working with people who understand these methodologies. When you move to the life sciences and medicine, these people aren’t brought up or educated to think there are predictive models. But that’s what we have to implement if we’re going to do personalized medicine—it’s all about saying something ahead of time so it’s actionable by clinicians. There’s quite a challenge to bring these people on board, and you can’t underestimate the importance of that. It does include education and training efforts to make them see the world in a similar way, and we’ve had to engage in those kinds of efforts to broaden the appeal of the methodology.

我認為，對於年輕一代的社群來說，讓他們接觸我們今天討論的技術，確實能激發他們的熱情。他們未來會期待在日常生活中使用這些技術。因此，看到整個技術全貌 (Whole Panoply) 非常重要。正如我之前提到的，引入不同學科的個人至關重要。大家都談到了這一點。我對複雜性 (Complexity) 很感興趣，我寫了一本書叫做《複雜性的前沿》(Frontiers of Complexity)。我們理解複雜性意味著系統的整體大於部分之和 (The Whole is Greater than the Sum of the Parts)。如果我們能在相同的波長上合作，擁有互補的專業知識 (Complementary Expertise)，我們能取得的成就將無窮無盡。這是你必須準備去做的事情。這種跨學科工作更具挑戰性，因為你不可避免地會走出自己早先設定的舒適區 (Comfort Zone)。但最終的回報遠遠更大，這是我希望傳達的啟發，特別是給那些剛踏上這條路的年輕人。

I’d say with the younger communities now, bringing them into contact with the technologies we’re talking about here really does start to fire them up. They’ll expect to use those in their daily lives in the future. So it’s really important to see the whole panoply there. As I was saying earlier, the ability to bring in individuals with different disciplines is vital. People have talked about this. I’m interested in complexity—I wrote a book called Frontiers of Complexity. We understand what we mean by that: systems where the whole is greater than the sum of the parts. That’s what we’re attempting to do with individuals—if we can work together on the same wavelength with different kinds of complementary expertise, there’s no end to the achievements we can make. But that’s what you have to be prepared to do. It’s more challenging because you inevitably step outside a comfort zone you may have created for yourself earlier. The rewards in the end are far greater—that’s the inspiration I’d hope to convey, especially to younger people just setting out on that pathway.