Cédric Villani

📌 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 commentaire

Allons-nous continuer la recherche scientifique ? Four years after winning the Fields Medal, Grothendieck gave a talk at CERN with the title "Are we going to continue scientific research?" At this stage of his life, Grothendieck was doing a balancing act between mathematics and ecological activism. One of his claims is that in some future, the ecological disaster will be so clear and impactful that scientific research will simply stop being a priority. His talk wasn't a call to stop research to do ecological activism. It was instead a premonition on why activism or survival will naturally take the energy and the minds of people who would otherwise do scientific research. Grothendieck's talk clearly anticipated today's dilemmas: "Has the Last Great Space Observatory Already Launched? Astronomy's future may be slipping away—one climate disaster at a time" https://lnkd.in/eMU25CMT Please listen to the talk https://lnkd.in/et2G-33a

10

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 commentaires

New Paper : Communities of innovation have a life cycle, right. Did you know that they attend places according to this cycle ? Here we take a detailed look at how video-mappers in Montreal attend places and what resources they draw from them. We apply an approach based on social network analysis to measure the collective and structural properties of innovation communities at each stage of the cycle. New recommendations are put forward for steering the development of creative and cultural ecosystems, based on a detailed understanding of how entrepreneurs move from one place to another. https://lnkd.in/ei3myMMb LEMNA - Laboratoire d’Economie et de Management Nantes-Atlantique IAE Nantes Nantes Université Samoa AIMS Stratégie Mosaic-HEC Montréal Heloise Berkowitz ANR SCAENA

28

I'll be giving a course on Multi-Agent Systems and Evolution with Tom Lenaerts and Marco Saponara in ESSAI 24 organised by European Association for Artificial Intelligence EurAI this July in Athens. If you are interested, you can register at https://lnkd.in/eEEm7jGf . Please share! #EGT #AI

20

The Institut Origines of Aix-Marseille University is recruting ! We are thrilled to announce an open PhD position focusing on The Atmospheric Composition of the Ice Giants Uranus and Neptune, in preparation for the next large international mission: The Uranus Orbiter and Probe mission (UOP). This thesis aims to model the abundance profiles of key volatile species in Uranus and Neptune, to ultimately help our team define, design and operate an atmospheric entry probe to be sent on these planets. Prepare for future explorations and contribute to our understanding of outer planets, enhancing our team’s capabilities in planetary science: https://lnkd.in/eMrK79Vz Image credits: K. Mandt (2023) DOI: 10.1126/science.ade8446 Institut Origines Aix-Marseille University CNRS European Space Agency - ESA NASA - National Aeronautics and Space Administration

91

2 commentaires

CAPITAL WORKSHOP: sCalable And PrecIse Timing AnaLysis for multicore platforms Friday June 14th. Toulouse and possible remote attendance with a keynote from Liliana Cucu-Grosjean, Inria Paris and StatInf "Statistical, stochastic or probabilistic (worst-case) execution time – what impact on the multicore time composability?" and 4 invited speakers: **Thomasz Kloda** (LAAS, Toulouse, France):       Memory-centric scheduling for phased execution models on multi-core platforms **Federico Aromolo** (Sant'Anna University, Pisa, Italy)       Timing Analysis of Parallel and Accelerated Software with Event-Driven Delay-Induced Tasks **Filippo Muzzini** (UniMORE, Modena, Italy)       Resource Contention Analysis in GPU-Accelerated Embedded Platforms **Frederic Bonamy** (Thales Avionics, Toulouse, France)       Multicore Challenges in Avionics Applications More details: https://lnkd.in/g2SeMArg Free (but mandatory) registration for in-person participation: https://lnkd.in/gNA2nbpu? Topic of the seminar: The design and the implementation of time-critical applications upon multi-core processors are considered. Multi-core processors have the potential to offer high computing power. Unfortunately, their extensive use of shared resources (e.g.,caches, DRAM, buses, etc.) makes the design and the implementation difficult to predict — especially in situations where hard real-time constraints must be satisfied. Recent works show that an important challenge is to design a precise and scalable timing analysis. To address this challenge the research community consider closely both sides of the system: software and hardware. The aim of this co-design approach is therefore to guarantee scalability, precision and low complexity of the solution without compromising flexible efficient use of resources. The solution must be tailored for the target application and platform. In particular by identifying the shared resources (memory, bus, cache) and by reducing the complexity based on interference delay using, for instance, a tailored mapping/scheduling, bandwidth regulator. The solution must be also based on a good use of the hardware architecture (memory banks, cache, communication media) with techniques like physical/temporal partitioning to obtain a precise solution. Organisers Thomas Carle (Université Toulouse III Paul Sabatier, IRIT, France) Eric Jenn (IRT Saint-Exupéry, Toulouse, France) Steering Commitee Thomas Carle (Université Toulouse III Paul Sabatier, IRIT, France) Eric Jenn (IRT Saint-Exupéry, Toulouse, France) Juan M Rivas (Universidad de Cantabria, Santander, Spain)

16

1 commentaire

Thursday at #ICLR2024 showcases a large chunk of our very diverse #realitycentricAI research agenda! You can find our students first at Poster Session 5 (10:45 CEST). The perfect opportunity to meet us and discuss our work. You can find all info about our early day papers below! #LLMs face a critical constraint: their limited “attention span”. This poses a challenge for generating cohesive outputs, particularly in lengthy tasks. In response, the researchers introduce #L2MAC, the first practical LLM-based general-purpose stored-program automatic computer (von Neumann architecture) framework, an LLM-based multi-agent system, for long and consistent output generation, setting a new state-of-the-art performance for generating large codebases and real-world applications for general-purpose extensive text-based tasks. "L2MAC Large Language Model Automatic Computer for Extensive Code Generation" by Samuel Holt, Max R., Mihaela van der Schaar Link: https://lnkd.in/eGEJV9Ny We have published a comprehensive companion piece on L2MAC here: https://lnkd.in/eEqpsknV   The lab’s researchers are driving forward personalised medicine through significant advancements in treatment effect estimation. In a ground-breaking paper, the researchers introduce a novel solution leveraging ordinary differential equations (#ODEs) to enhance interpretability and address irregular sampling challenges. By departing from traditional neural network-based inference methods, the research offers insights into interpretable models that bridge the gap between treatment effect estimation and equation discovery – a world’s first and significant paradigm shift. "ODE Discovery for Longitudinal Heterogeneous Treatment Effects Inference" by Krzysztof Kacprzyk*, Samuel Holt*, Jeroen Berrevoets*, Zhaozhi Q., Mihaela van der Schaar Link: https://lnkd.in/eWcUADiT Have a look at our tutorial series on #TreatmentEffectInference: https://lnkd.in/eDp2Z-8w   The lab’s contributions extend to integrating the capabilities of LLMs within Bayesian Optimization (BO) with the introduction of LLAMBO. BO is commonly used to optimize expensive, black-box functions, including in robotics and experimental design, and most prominently for autoML! LLAMBO methodology frames each of the optimisation components of BO into natural language. This enables them to take advantage of the prior knowledge and few-shot learning capabilities of LLMs to iteratively propose and evaluate promising solutions conditioned on historical evaluations and contextual information. LLAMBO offers a versatile solution that seamlessly integrates LLM capabilities into the BO acquisition process, demonstrating the transformative potential of AI-driven methodologies. "LLMs to Enhance Bayesian Optimization" by Tennison Liu*, Nicolás Astorga*, Nabeel Seedat, Mihaela van der Schaar Link: https://lnkd.in/eA6hxmCw Learn more about #AutoML: https://lnkd.in/eADPqxKa

12

#EHDS and #AI: what implementation challenges can be expected? That's what we discussed this week at the Université Toulouse 1 Capitole, at a brilliantly organised conference on the future challenges posed in part by the adoption of EHDS. Thanks again to Nathalie De Grove Valdeyron for inviting me on behalf of Sciensano and organising this wonderful event. I was able to revisit the Healthy Data consultation (https://lnkd.in/eSRNACVb) and re-emphasise the need to co-construct a shared vision of the pursuit of societal benefits that should guide the secondary use of health data, and the need to provide ongoing opportunities for public engagement in developing this future framework.

63

2 commentaires

Save the date! I look forward to see you in this Webinar, to discuss about this important topic of Speeding!

9

The first private 5G network of its kind for a French University is finally deployed.

21

The #AIActGame, a good way to learn in an easy and fun way what the #AIACt, the new #European regulation on #ArtificialIntelligence, is about. Designed by Thomas Le Goff, PhD from Télécom Paris Institut Polytechnique de Paris, with Winston Maxwell, professors and researchers specialised in digital law I tried the game and recommend it!

4

1 commentaire

[New Research Publication] We are thrilled to announce that our latest research article has been accepted in the proceedings of the 50th Euromicro Conference Series on Software Engineering and Advanced Applications (SEAA), to be held in Paris, France from 28-30 August. This article is co-authored by Maha Sroor, Tommi Mikkonen and Teerath Das (University of Jyväskylä), and Sandun Dasanayake (University of Oulu), Finland. In our paper, titled "Practitioners' Perceptions of Security Issues in Software Containers: A Qualitative Study," we investigate the critical topic of container security from the practitioners’ point of view. We conducted semi-structured interviews with SE practitioners across various domains to gain insights into their real-world experiences and strategies for managing container security. Our analysis identified emergent patterns and relationships between major themes, offering a comprehensive model of container security practices. Our findings aim to bridge the gap between academic research and industry practices, providing valuable insights for both communities. We are looking forward to sharing our research and engaging with fellow experts at SEAA 2024! A Thank you to everyone who supported us on this journey and Value Creation for Cyber-Physical Systems and Services (CPSS) Group. #SEAA2024 #SoftwareEngineering #ContainerSecurity #Research #Cybersecurity #TechInnovation

25

1 commentaire

[✉️🖋️📝] #IQSOG #XerfiCanal #Juillet2024 Dr Maria Mercanti-Guérin / Ph.D., associate professor at IAE Paris - Sorbonne Business School, discusses her book "Web Crash: The day the Internet collapsed." She explores the collapse of the commercial web, characterized by a model centered on buying and advertising. Mercanti-Guérin uses a generative AI named Lumaine to imagine various post-collapse scenarios with different probabilities of occurrence in the short, medium, and long term. The first scenario, called "Web ICO" (brain-computer interface), envisions a future where computers and mobile devices are replaced by chips implanted in the brain, allowing direct interaction with the cloud. This dystopian scenario, while alarming, has a high probability of occurring in the long term according to Lumaine, due to significant investments by major tech companies in the health sector. The second scenario is the "Web Amazon," where a giant platform dominates the web, creating a digital monopoly. However, Lumaine estimates that legislation and consumer resistance will prevent such a monopoly from emerging, giving this scenario a low probability of occurrence. Mercanti-Guérin notes that meta-applications, like those envisioned by Elon Musk with X, could follow this path. The third type of web is the "Ubiquitous Web" (Ubiq), where the web is omnipresent in all media, such as connected TVs and digital displays. Lumaine considers this scenario to have an extremely high probability of occurring in the short term, as it is already in progress. The fourth scenario is the "Decentralized Web" or "Web 3.0," focused on the metaverse and utilizing virtual reality headsets, cryptocurrencies, and digital wallets. This web, although currently in crisis and accumulating losses, has a high probability of occurring in the long term according to Lumaine. The last two scenarios include the "Authoritarian Web," where government control is strengthened, particularly in China with strict internet usage restrictions, and the "Militant Web," where subscriptions and activists fund an independent web free from advertising. The authoritarian web has a high probability of occurring today and in the future, while the militant web has a low probability of occurrence. Mercanti-Guérin concludes by noting that some of these scenarios could combine. For example, a web governed by AI could align with an authoritarian web, while the militant web could ally with the decentralized web, sharing the philosophy of web decentralization. She highlights that using a generative AI like Lumaine to develop these scenarios demonstrates its potential as a super assistant for understanding and anticipating the web's evolution. Source: Mercanti-Guérin, M. (2023). Web Crash: The day the Internet collapsed. EMS Editions. https://lnkd.in/eX7Hagvw

5

----------English version of previous post   And another business sector! The space industry is leading the way in integrating environmental considerations. Many challenges are on the agenda, including major issues such as space pollution and the danger of space debris and systems re-entering the atmosphere.   R&T challenges are one-off calls for ideas organised to complement the CNES R&T system on technical subjects of interest. The aim of these challenges is to stimulate the emergence of innovative solutions for the space sector, with an opening to new players, whose development can be pursued via Research and Technology.   "Taking environmental issues into account is crucial. Through this challenge, CNES aims to give a new momentum to innovative solutions that will drastically reduce the environmental impact of the design, development and operation of space systems. As this is an issue common to both orbital systems and space transportation, this challenge is being co-organised for the first time by these two teams working to prepare for the future. We should also highlight the substantial budget allocated by CNES, which is commensurate with the challenges and ambitions in this area", say Céline Angélélis, Deputy Director for the Preparation of the Future of Orbital Systems and Ursula Aniakou, Deputy Director for Future Projects and Innovation in Space Transport Systems.   A congratulations to Airbus Defense & Space, Thales Alenia Space, Mecano-ID and IRT Saint Exupéry who are entering into a contract with CNES to start their developments to face the current climatic and environmental challenges.   Bamboo will soon be taking its first trip into space 😉   #space #sustainability #pollution #BambooisGrowing #1000solutions              https://lnkd.in/eJ5eST3V

6

All our planned project deliverables were accepted by the #europeancommision following our 1st review meeting! Congratulations to the whole Neuropuls consortium! Check out our latest public deliverables below available in #Zenodo! #artificialintelligence #photonics #security #neuromorphic #horizoneurope

17

We are thrilled to announce our 2024 individual grant research recipients! Congratulations to all the winners: Marco De Virgilis and Daniel Lupton : Applications of Graph Theory in Territorial Ratemaking Himchan Jeong and Bin Zou: Development of Telematics Risk Scores in Accordance with Regulatory Compliance Linfeng Zhang and Zhiyu Quan: Entity-Specific Cyber Risk Assessment Using Insurtech-Empowered Risk Factors and State-of-the-Art Machine Learning Algorithms Yang Lu and Hanieh Amjadian: Evaluating Model Risk in Property and Casualty Insurance Tong Zhiwei and Haibo Liu: Integrated Frequency and Severity Modeling for Extreme Losses Fei Huang and Joshua Loftus: Interpretable and Fair Insurance Risk Pricing Using Causal Models Liang (Jason) Hong, PhD, FSA: A New Approach to Detecting Insurance Fraud Kun Shi: Unveiling Nuances in Categorical Variables to Enhance Information Integration in Actuarial Applications Selected from nearly 50 proposals, these groundbreaking projects reflect the forefront of actuarial research, addressing key topics like machine learning and cybersecurity. "Investing in these projects supports CAS members in developing new skills, learning new techniques, and ultimately enhancing their professional capabilities,” says CAS Vice President-Research and Development, Morgan Bugbee, FCAS. Stay tuned for the announcement of the 2025 Individual Research Grants program this fall! #CASResearch #ActuarialScience #GrantWinners #InnovationInInsurance https://lnkd.in/eUnk5NNk

102

🎉I am very happy to share our publication co-authored with Lionel Nicod and Sylvie Llosa. This article was recently published in the Journal of Business Research and it is the collaborative work of a team of Aix-Marseille Université researchers (from FEG - Faculté d'économie et de Gestion Aix-Marseille Université, IUT, and IAE Aix-Marseille Graduate School of Management), who are all part of the CERGAM - Centre d'Etudes et de Recherche en Gestion d'Aix-Marseille research laboratory. 📚In this empirical study, we investigate the mitigation of bypass on #sharingplatforms, and whether message framing is effective in reducing bypass. In particular, we address the following research questions: 🎯Is it possible to reduce bypass by displaying a persuasive message? 🎯Are both approaches—a positively framed and a negatively framed message—effective in reducing bypass intention? Is one approach more effective than the other? 🎯What are the underlying mechanisms explaining the effect of a persuasive message on bypass intention? ✨If you want to know more, the full article is now available for reading and download here: https://lnkd.in/ee4CkcJT 🙏Many thanks to the guest editors Professor Schaefers, Professor Leban, and Professor Woisetschläger for their guidance. We also thank the three anonymous reviewers for their valuable feedback and suggestions that significantly strengthened our paper. #jbr #JournalofBusinessResearch #AMU #FEG #IAE #bypass

83

7 commentaires

[📢Parlons IA] Notre sélection hebdomadaire d'articles sur l'#IA & la #Data, c'est ici tous les vendredis ! ⬇️ Learning how to use AI could boost your pay by 25%, study finds > https://hubs.ly/Q02yfbJL0 Revolutionary AI Device Mimics Human Brain With Few-Molecule Computing > https://hubs.ly/Q02yf9Cq0 AI Uncovers Hidden Differences in Male and Female Brain Structures > https://hubs.ly/Q02yf8jr0 Airline to 'better manage' flights with AI use > https://hubs.ly/Q02yf8RY0 AI will be everywhere. How should we prepare for the ethics around this? > https://hubs.ly/Q02yfc6K0 Predicting Chaos With AI: The New Frontier in Autonomous Control > https://hubs.ly/Q02yf8SY0 An AI tool that can classify brain tumors into 10 major subtypes with 95% accuracy > https://hubs.ly/Q02yf7_20 World is ill-prepared for breakthroughs in AI, say experts > https://hubs.ly/Q02yf8l70 No One Truly Knows How AI Systems Work. A New Discovery Could Change That > https://hubs.ly/Q02yfch10 New Study Finds Age and Race Impact Artificial Intelligence Performance on Digital Mammograms > https://hubs.ly/Q02yfc4-0

22