Stefan Güldenberg

CROSS Call 2025 offers funding to interdisciplinary research endeavors addressing relevant issues on society and technology - exclusively for teams integrating researchers from EPFL and University of Lausanne - UNIL Open to all topics and disciplines. Submit your project until 27 June! #researchfunding #interdisciplinaryresearch #societalchallenges #technology #EPFL #UNIL Details here👇🏼

12

It took a bit of time, but our paper (Damien Brun, Sébastien George and I) on the design and evaluation process of the Keycube, a cubic keyboard for versatile text input, is published in the renowned Taylor&Francis Human-Computer Interaction Journal ! 😊 For access to the paper : https://lnkd.in/eP9jEP2Y #hci #textinputdevice #keyboard #mixedreality #virtualreality

46

2 Comments

📌 NEW Training Session📌 Topic: Social Signals in Affective Cognitive Sciences: Recognition and Production with Ali Oker, Université de Reims Champagne-Ardenne In this talk, we will present the theoretical and methodological contributions of computer science to the study of the human mind in cognitive sciences. Cognitive science is defined as a cluster of interdisciplinary domains (such as psychology, computer science, neuroscience, linguistics, anthropology) that examine how a 'system' can learn, transform, and produce a response during information processing. Particularly in what we call "affective cognitive science," in psychiatry and social neuroscience, advancements in both hardware and software development have enabled scientists to conduct more precise and interesting experimental work. More information: https://lnkd.in/dqruMrAn 🇪🇺 Funded by the European Union. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or European Research Executive Agency (REA). Neither the European Union nor the granting authority can be held responsible for them.

13

1 Comment

Per Capita Publications (2023) I just created a map for teaching purposes. The map depicts per capita Scopus-indexed publications in 2023, divided into quartiles. Some interesting observations: The two highest-scoring countries (Switzerland and Denmark) produce three times as many publications per head as the United States, which in turn churns out three times as much as China on a per capita basis. To me, the biggest surprise was that South Korea (1,958 publications per million inhabitants), Taiwan (1,856), and Malaysia (1,331) have all overtaken Japan (1,042). Also, it may surprise some people that the United States is the worst-performing country in the Anglosphere (Australia: 4530; Ireland: 4203; UK: 3522; New Zealand: 3501; Canada: 3314; U.S.: 2101).

33

5 Comments

#GPT4 | 𝗔𝗜'𝘀 𝗡𝗲𝘄 𝗣𝗼𝘄𝗲𝗿 𝗼𝗳 𝗣𝗲𝗿𝘀𝘂𝗮𝘀𝗶𝗼𝗻: 𝗜𝘁 𝗖𝗮𝗻 𝗖𝗵𝗮𝗻𝗴𝗲 𝘆𝗼𝘂𝗿 𝗠𝗶𝗻𝗱 | A team at EPFL, led by Prof. Dr. Robert West, has examined the persuasive capabilities of large language models (LLMs), specifically OpenAI's GPT-4. The study aimed to compare the persuasive abilities of AI models to those of humans, and understand the role of personalization in AI's power of persuasion. The study placed 820 participants in debates against either a human or an AI chatbot, with or without personal data available to the debater. Findings revealed that GPT-4 was most persuasive when it had access to personal information. This groundbreaking research not only emphasizes the potential influence of LLMs on human behavior but also underscores the essential need for further research to understand and regulate the persuasive power of large language models. 👉 Learn more >> https://lnkd.in/dKv9EZDd 🇨🇭 Follow #ScienceSwitzerland for the latest news and emerging trends on Swiss science, technology, education, and innovation >> www.swissinnovation.org Follow us >> Science-Switzerland #Science | #Education | #Research | #Innovation

4

SimBricks started out four years ago as a research project, during turbulent times for everyone (pandemic & see picture). It all started with casual discussions between Antoine Kaufmann and Jialin Li on the pains of evaluating our heterogeneous systems research, and two weeks later a first workshop vision paper submitted (but sadly not accepted) to HotNets'20. Hejing Li joined the project a few weeks later. Fast-forward four years (and another by now grown-up puppy), the research project has grown to a revolving team of 8-10 people, a substantial self-contained open-source prototype actively used as a tool in multiple other projects, and a handful of published and in-progress papers. On top of all of this, part of the team (Hejing, Jonas, Jakob, Marvin, and Antoine) is now exploring commercializing some of our ideas. #FullSystemSimulation #Research #TechTransfer #Startup

27

2 Comments

May 15-17 2024: The 6th edition of "Journées Approximation" at the University of Lille Université de Lille The conference featured excellent talks on mathematics and its applications, fostering valuable exchanges among researchers in approximation theory and related fields. https://bit.ly/3yAuORK

3

OpenAI has now made its GPT store free to use. A GPT is a customized chatbot, based on GPT-4o. I've trained the TeachSmart GPT on the content of my book "Practical Pedagogy: 40 New Ways to Teach and Learn", plus other material to try and keep its responses focused and accurate. It's designed for teachers, researchers and EdTech designers to ask questions about effective ways to teach, innovative methods of teaching and learning, learning design, and educational technology. For example, try asking: "What is Teachback and how can I use it in my classroom?", "How can I design an engaging 30 minute lesson to teach Spanish to young children?", "Does a flipped classroom approach work for fully online learning?", "Should I match my lesson to students' learning styles?". You can see below the start of its reply to the last question. Give it a try! To access the GPT store you need to got to https://chatgpt.com/ then register with an email and password and give your date of birth. Then click "Explore GPTs" and search for TeachSmart. Do let me know about any errors or weaknesses and I'll do my best to correct them. #aied #ChatGPT #GenAI #OpenAI

222

21 Comments

Happy to invite you to a talk by Dr. Amro Najjar Luxembourg Institute of Science and Technology (LIST) Talk Title: Interactive and Embodied Explainable AI. Abstract: Over the past five years, the field of Explainable AI (XAI) has received a growing attention. Initially aiming to investigate the intriguing results of some black-box machine learning mechanisms such as DNN, XAI research has later expanded to several other domains such as explainable planning, explainable recommender systems and explainable agents. Central to this research is the development of human-centered explanations, prioritizing interactive communication over static methods. Compared to static explanations, engaging the user in interactive (back-and-forth) communication is crucial as it allows the user to build a more solid and backed-up knowledge and awareness of the subject being discussed with the AI system. This helps facilitate a deeper understanding by allowing real-time clarification of doubts, addressing misunderstandings, and responding to follow-up inquiries. Furthermore, integrating these interactive techniques with embodied XAI—where agents leverage non-verbal cues such as facial expressions and gestures—has proven exceptionally effective. Such embodiments enhance the clarity of communication and boost user acceptance and trust. This talk explores cutting-edge developments in Interactive and Embodied XAI, spotlight key applications, and delve into a case study on an Interactive Explanation-based Nutrition Virtual Coaching System, illustrating how these advanced techniques are reshaping user interactions with AI systems. For more information about the talk timing and location; please visit the blow website https://lnkd.in/duSBqXPX

17

Strategic success of a Polish start-up. Only 2 start-ups from Poland are among the top 103 from around the world.  MassChallenge is the strongest promoter of innovation for bold entrepreneurs. Way to go AlphaPowders! #AlphaPowders

4

1 Comment

I am happy to share with my network and academic colleagues our newest scientific article recently accepted at INFORMS Journal on Computing, co-authored with my colleagues and friends Marilène Cherkesly and Matthieu Gruson, on the discrete ordered median problem. We propose a novel decomposition scheme to address the DOMP that breaks the ranking attribute of the problem to solve multiple easier problems cast as simpler covering or packing problems. You can access the article on IJOC's website using the following link https://lnkd.in/eeq_5qc4, or you may simply drop any of us an email if you encounter issues accessing it. We will be happy to send it to you. In addition, we have released a Julia package of our method. Since the package is registered you simply need to add it by typing "add DiscreteOrderedMedian" within the package REPL. The code is open-sourced using a MIT license and available at https://lnkd.in/eVpWBYdA.

152

3 Comments

#ComputerScience | 𝗛𝗼𝘄 𝗔𝗜 𝗧𝗮𝗰𝗸𝗹𝗲𝘀 𝗖𝗼𝗺𝗽𝗹𝗲𝘅 𝗗𝗮𝘁𝗮 𝗦𝗮𝗺𝗽𝗹𝗶𝗻𝗴 | A team of scientists led by Prof. Dr. Florent Krzakala and Prof. Dr. Lenka Zdeborova at EPFL have made a significant breakthrough in understanding the performance of neural network-based generative models compared to traditional data sampling techniques. This study focuses on a specific class of probability distributions related to spin glass materials and statistical inference problems. The research drew inspiration from the intricate world of spin glasses to analyze modern data generation techniques, shedding light on how neural network-based models navigate complex data landscapes. The study found that diffusion-based methods might face challenges due to a first-order phase transition, thereby revealing the efficiency and limitations of modern generative models. This breakthrough is a crucial step towards enhancing our theoretical understanding of generative models. 👉 Learn more >> https://lnkd.in/ehmswBxQ 👉 Original publication >> https://lnkd.in/ggSq_Qnn Image: EPFL / iStock 🇨🇭 Follow #ScienceSwitzerland for the latest news and emerging trends on Swiss science, technology, education, and innovation >> swissinnovation.org Follow us >> Science-Switzerland #Science | #Education | #Research | #Innovation

4

Dear All, After the 9th rTAIM Online Seminar (available here: https://lnkd.in/dCvzV6sb), we are happy to announce the 10th rTAIM Online Seminar with the participation of Professor Edmund Ugar (University of Johannesburg, South Africa), on the 27th May 2024, 15h-16h00 Lisbon Time Zone, via Zoom. Title: In Light of Medical Machine Learning Technologies in Sub-Saharan Africa: African Agency Reconsidered Abstract: The growing application of machine learning techniques in the second decade of the present century has made artificial intelligence to play a more significant role in medicine and, increasingly, in public health. Image recognition, in particular, has become extremely effective, and clinicians increasingly rely on machine learning technologies for clinical diagnosis and prognosis of medical conditions. While these technologies have proven to be relevant and efficient in carrying out medical diagnosis, in this talk, I argue that they threaten an integral aspect of what constitutes human agency for Africans – interpersonal relationships. Overreliance on machine learning technology for clinical diagnosis and recommendations may diminish the values of interpersonal relationships of identity and solidarity between patients and doctors in the sub-Sahel. I show how vital these normative values are in the making of a person and their agency within the sub-Saharan African value system. Short Bio: Edmund Terem Ugar is a doctoral candidate in Philosophy at the University of Johannesburg, a research fellow at the Centre for the Philosophy of Epidemiology, Medicine and Public Health at Durham University, and a researcher at the University of Johannesburg Centre for Africa-China Studies. His research lies broadly within the Philosophy of Social Technologies within Medical and Healthcare Domains. Zoom: Link: https://lnkd.in/dsYfwEE2 ID: 943 0582 3062 Password: 351181 More info: https://lnkd.in/dAeX8JmK

3

My chapter on generative AI in tertiary education has just been published. Such a nice gift from Springer today, a month before the Human-Computer Interaction International (HCII 2024), where I have the privilege, entrusted to me by Program Chairs Adela Coman and Simona Vasilache, of being a Program Board member and a session chair for the SCSM https://lnkd.in/dGgFX-EC. #hcii #hcii2024 #scsm #socialcomputing #socialmedia #humancomputerinteraction #ai #generativeai #aiineducation #aiintertiaryeducation #aiinhumanities #aiinculturestudies

38

6 Comments

My new article titled "The synchronic and diachronic evolution of key themes around SDG 4 before and after 2015: From a Quantitative analysis of web‑downloaded texts" is published in the International Review of Education journal. This paper presents a new potential of critical discourse analysis using quantitative text mining of web-posted documents to identify the thematic trends before and after SDG adoption in 2015. The paper focuses on the global discourse related to the education goal of SDGs (SDG4), but it provides insights not only for educationists but also for people who are interested in the evolution and prospect of this international agenda. https://lnkd.in/gRs7jzVw #critical_discourse_analysis #text_mining #SDG4 #learning_outcomes #COVID_effects

20

1 Comment

Another new development in the education sector! chatGPT introduces chatEdu for universities. Will the initial adaptation be advantageous? Time will tell. However, the idea looks very promising, and it seems like the whole education system may experience revolutionary change. #Innovation #Education #chatGPT #chatEdu Read more about this development here: https://lnkd.in/dnmDhfRa

3

🎓 My PhD thesis is now available online on Biblioteche dell'Università della Svizzera italiana. 📚 You can find it using the following link: https://lnkd.in/eJChp7Sb #TestCaseGeneration #FaultLocalization #Debugging #DataScience #Python #NeuralNetworks #DeepLearning

40

2 Comments

We are pleased to announce the publication of our new book chapter "The Role of Motivation and Academic Preparation for the Students’ Entrepreneurial Potential" in the book Entrepreneurial Motivations Strategies, Opportunities and Decisions edited by Marco Valeri. #entrepreneurship #motivation #students #entrepreneurial #potential #academic #preparation #HEI #universities #Portugal

19

3 Comments

"Sapere aude: ne non te ratio subiuget priusquam metus." "Dare to know: so that fear may not subdue you more than reason does." I am so happy to share that I have been awarded the Independent Research Fund Denmark 'Sapere Aude' starting grant! We will explore the roles of phosphorylation in the disordered domains of the transcription factor FoxO3, critical and aging and cancer, using NMR and complimentary techniques. Stay tuned for open Postdoc and PhD positions in the near future! https://lnkd.in/dcgM4Pum #SapereAude #DFF #nmr #disorderedproteins #phosphorylation #foxo3

98

8 Comments