#École polytechnique fédérale de Lausanne

[younes-ben-amara]

المهارة الرئيسية التي يتّسم بها جميع الفاعلين الناجحين في البحث العلميِّ [هي الفضول!]

[nerdwelt]

Staatliche Regulierung kann die Gefahren sozialer Medien wirksam eindämmen

Neue Forschungsergebnisse zeigen, dass staatliche Gesetze zur Kennzeichnung und Moderation gefährlicher Inhalte in sozialen Medien wirksam zur Schadensminderung beitragen können, selbst auf schnelllebigen Plattformen wie X (ehemals Twitter). Social-Media-Beiträge, die Terrorismus und Hass fördern, gefährliche Herausforderungen darstellen, die das Leben von Teenagern gefährden, oder solche, die…

[bpod-bpod]

Tumour Tracked

Tracking emergent colon tumours in real-time at the single-cell level over several weeks in a 3D lab-grown organoid system

Read the published research article here

Video from work by L. Francisco Lorenzo-Martín and Tania Hübscher, and colleagues

Laboratory of Stem Cell Bioengineering, Institute of Bioengineering, School of Life Sciences and School of Engineering, École Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland

Video originally published with a Creative Commons Attribution 4.0 International (CC BY 4.0)

Published in Nature, April 2024

You can also follow BPoD on Instagram, Twitter and Facebook

[materialsscienceandengineering]

Unlocking the magnetic superpowers of topological magnons

In the ever-evolving landscape of condensed matter physics, a recent breakthrough has emerged from the collaborative efforts of researchers at the Peter Grünberg Institute (PGI-1), École Polytechnique Fédérale de Lausanne, Paul Scherrer Institut in Switzerland, and the Jülich Centre for Neutron Science (JCNS). This synergetic work, driven by the trio Manuel dos Santos Dias, Nikolaos Biniskos, and Flaviano dos Santos and led by Stefan Blügel, Thomas Brückel, and Samir Lounis, has delved into unexplored magnonic properties within Mn5Ge3, a three-dimensional ferromagnetic material. Topology, a concept pivotal in contemporary physics, has already played a transformative role in understanding electrons in solids. From quantum Hall effects to topological insulators, the influence of topology is far-reaching. In this context, the focus has shifted to magnons—collective precession of magnetic moments—as potential carriers of topological effects. Magnons, being bosons, can exhibit unique phenomena akin to their fermionic counterparts.

Read more.

[moonwatchuniverse]

45 years ago… first Swiss astronaut Claude Nicollier In 1966 Claude Nicollier became a pilot in the Swiss Air Force, switching as an airline pilot with Swiss Air by 1974. In 1978, combining this with studies in Astrophysics, Nicollier was selected a member of the first group of European Space Agency astronauts. Between 1992 & 1999 the first Swiss astronaut flew 4 space shuttle missions and maintained his role as lead ESA astronaut up to 2007. February 1996, during his third mission, STS-75, Nicollier wore the Omega Speedmaster X-33 analog-digital chronograph prototype, which was officially launched as an operational astronaut/cosmonaut toolwatch by March 1998. Official 1999 ESA/NASA portraits show him wearing a Breitling Aerospace chronograph but in May 2005, Nicollier became member of the Board of Directors of the Swatch Group and full-time Professor at the École Polytechnique Fédérale de Lausanne. (Photo: ESA/NASA)      

[jaspersboy]

A team from École Polytechnique Fédérale de Lausanne in Switzerland made the breakthrough using specially designed photosensitizer dye molecules that when combined are capable of harvesting light from across the entire visible light spectrum.

The transparent properties of DSCs make them suitable for use in windows, greenhouses and glass facades, the researchers said, as well as in the screens of portable electronic devices.

They are also flexible, relatively low-cost and can be made using conventional roll-printing techniques. Theoretically, the price/performance ratio is also good enough to allow them to compete with fossil fuel electrical generation.

[normally0]

Obituary: Paul Chemetov (1928–2024)

Paul Chemetov, a giant in French architecture and urban planning, passed away in Paris on June 16, 2024, at 95. Known for designing France's Ministry of Economy and renovating the Grand Gallery of Evolution, Chemetov was a committed and cultured figure. Awarded Commander of the Légion d'honneur in 2016, he also taught at École des Ponts ParisTech and École Polytechnique Fédérale de Lausanne. He is survived by his son Alexandre, and daughters Marianne and Agnès.

[levysoft]

Una nuova ricerca mostra che i grandi modelli linguistici (LLM) come il GPT-4 possono sovraperformare significativamente gli esseri umani in termini di persuasività in situazioni di dibattito faccia a faccia.

In uno studio controllato, i ricercatori della École Polytechnique Fédérale de Lausanne (EPFL) e della Fondazione italiana Bruno Kessler hanno studiato il potere persuasivo dei grandi modelli linguistici come il GPT-4 in confronto diretto con gli esseri umani.

I partecipanti sono stati divisi in modo casuale in diversi gruppi e hanno discusso argomenti controversi. Il team di ricerca ha testato quattro diverse situazioni di dibattito: umano contro umano, umano contro IA, umano contro umano contro umano con personalizzazione e modello di intelligenza artificiale umano contro personalizzato.

Nella versione personalizzata, i dibattitori hanno anche avuto accesso a informazioni di base anonime sui loro avversari.

Il risultato: GPT-4 con accesso alle informazioni personali è stato in grado di aumentare l'accordo dei partecipanti con le argomentazioni dei loro avversari di un notevole 81,7% rispetto ai dibattiti tra esseri umani.

Senza personalizzazione, il vantaggio di GPT-4 rispetto agli esseri umani era ancora positivo al 21,3%, ma non statisticamente significativo.

I ricercatori attribuiscono questo vantaggio persuasivo dell'IA personalizzata al fatto che il modello linguistico utilizza abilmente le informazioni del profilo del partecipante per formulare argomenti su misura e persuasivi.

Secondo i ricercatori, è preoccupante che lo studio abbia utilizzato solo dati di base rudimentali per la personalizzazione - eppure la persuasività di GPT-4 era già così significativa.

Gli attori malintenzionati potrebbero generare profili utente ancora più dettagliati da tracce digitali, come l'attività sui social media o il comportamento di acquisto, per migliorare ulteriormente il potere persuasivo dei loro chatbot AI. Lo studio suggerisce che tali strategie di persuasione guidate dall'IA potrebbero avere un impatto importante in ambienti online sensibili come i social media.

I ricercatori raccomandano vivamente che gli operatori di piattaforme online adottino misure per contrastare la diffusione di tali strategie di persuasione guidate dall'IA. Una possibilità sarebbe quella di utilizzare sistemi di intelligenza artificiale personalizzati in modo simile che contrastano la disinformazione con controargomentazioni basate sui fatti.

I limiti dello studio includono l'assegnazione casuale dei partecipanti a posizioni pro o contro, indipendentemente dalle loro opinioni precedenti, e il formato strutturale predeterminato dei dibattiti, che differisce dalle dinamiche delle discussioni online spontanee.

Un'altra limitazione è il limite di tempo, che può potenzialmente limitare la creatività e la persuasività dei partecipanti, specialmente nella condizione di personalizzazione in cui i partecipanti devono elaborare ulteriori informazioni.

Lo studio completo è stato condotto tra dicembre 2023 e febbraio 2024 e finanziato dalla Swiss National Science Foundation e dall'Unione europea. È pubblicato come prestampa su arXiv.

Il CEO di OpenAI Sam Altman ha recentemente avvertito del potere persuasivo sovrumano dei grandi modelli linguistici: "Mi aspetto che l'IA sia capace di persuasione sovrumana ben prima che sia sovrumana nell'intelligenza generale, il che può portare ad alcuni risultati molto strani".

[thierrylidolff]

MÉDIAS SOCIAUX : PRIORITÉ AUX MODÉRATEURS HUMAINS ?

ARTICLE Contenus haineux en ligne : oui, les modérateurs humains ont les moyens de l’emporter Publié: 21 novembre 2023, THE Conversation Marian-Andrei Rizoiu, University of Technology Sydney, Philipp Schneider, EPFL – École Polytechnique Fédérale de Lausanne – Swiss Federal Institute of Technology in Lausanne Les réseaux sociaux sont devenus les « places publiques numériques » de notre…

[cumulations]

Resonance for August 14, 2023

Highlights of the Week

Welcome! This week we delve into the use of tools in applications of Large Language Models (LLMs). But that happens only after looking at a keynote by Yann LeCun, in which he critically analyzes the current state of LLMs and suggests additional model types necessary for real progress.

While ‘tools’ and ‘agents’ are often referenced together, it’s crucial to distinguish between them. When discussing LLMs, ‘tools’ often refer to software, applications, and resources that aid in creating, managing, deploying, or enhancing these models. This includes programming languages, libraries, platforms, APIs, among others. ‘Agents’, on the other hand, are the entities that interact with the models.

Next, we explore “Flows”, a groundbreaking AI framework from researchers at the École Polytechnique Fédérale de Lausanne (EPFL) and Paris Sciences et Lettres University (PSL University). This approach promises advances in AI system functionality by simplifying the composition of various models and tool linkages.

We then investigate how LLMs can utilize self-supervised learning to enhance their functionality, introducing the Toolformer project and its potential synergy with tools like the previously reviewed Gorilla project.

Finally, we discuss democratizing Reinforcement Learning with Human Feedback and the role of DeepSpeed-Chat in making it more accessible and affordable. Tools await us!

Towards Machines that can Learn, Reason , and Plan

To start us off, I want to draw your attention to a Keynote given by Yann LeCun at the “Impact of chatGPT” talks on July 21, 2023. The address, entitled “Towards Machines that can Learn, Reason, and Plan”, is one one of the best I’ve listened to. He notes the shortcomings of the spate of LLMs now getting big attention, and points out the likely sources of data and AI technologies which, when used in conjunction with the LLMs, could possibly address the problems. The address is available in this video recording . The slides are available as a .pdf here: Objective-Driven AI

According to LeCun, Auto-Regressive Generative Models suck. (His words. Not mine). He goes on to state that what we need are the technologies and systems that address three challenges:

Learning about and learning to use representations and predictive models of the world.

Learning to reason.

Learning to plan complex actions which satisfy objectives.

In addition to the critique of LLMs, he touches on the importance of open source AI, hybrid systems for reasoning and planning. It’s well worth your time.

Reasoning and Collaborating AI

Right after listening to Yann LeCun’s address, I came upon Flows: Building Blocks of Reasoning and Collaborating AI. It was almost as if, in response to one of the key messages of the address, this project from École Polytechnique Fédérale de Lausanne and Paris Sciences et Lettres University appeared magically.

Imagine an AI system that’s like a Lego set, with parts that can be assembled, disassembled, and reassembled in various ways to create different structures. This is the concept behind “Flows,” a new AI framework presented in this paper. Flows are like individual building blocks of computation that can communicate with each other. These blocks can be combined in numerous ways to model complex interactions among multiple AI systems and humans. The beauty of Flows is that they reduce complexity by breaking down big tasks into smaller, manageable parts.

As a proof of concept, the researchers used Flows to improve the performance of AI in competitive coding, a task that many advanced AI models find challenging. The result was a significant improvement in the AI’s proficiency. To make this new framework accessible for further research, the authors have introduced the aiFlows library, a collection of Flows that researchers can use and build upon.

How an LLM Might Use Self-supervised Learning About How to Use Tools

As you might recall, one of last week’s recommended readings was Gorilla: Large Language Model Connected with Massive APIs, the open source project which identified the best APIs to be used by LLMs for specific purposes, and guidance about how they might address them. But in this configuration, it’s not clear that an LLM will already have the skills to follow this guidance. That’s where projects and offerings like Toolformer, a language model that can teach itself to use tools, provide the potential solution.

The authors note that large language models (LLMs) have become incredibly popular mainly because of their outstanding performance on a range of natural language processing tasks. One of their most significant differentiating factors is their impressive ability to solve new tasks from just a few examples or text prompts. This makes it all the more puzzling that these ostensibly all-knowing LLMs frequently have difficulties with fundamental functions like executing arithmetic operations or with being able to access up-to-date information. At the same time, much simpler and smaller models perform remarkably well in this space. The work of researchers from Meta AI Research and Universitat Pompeu Fabra reports that Toolformer not only decides which APIs to call, when to call them and what arguments to pass, but it comes by this knowledge and skill by ‘self-supervised learning, requiring nothing more than a handful of demonstrations for each API.’ It would seem that the combination of Gorilla and Toolformer might well be a way forward.

ALL Model Learning May Not Be Self-supervised.

Reinforcement Learning with Human Feedback (RLHF) is a method where an artificial intelligence system learns to improve its actions or decisions based on feedback it receives from humans. DeepSpeed-Chat is a novel system designed to make RLHF training for powerful AI models more readily and economically available. With easy-to-use training, a scalable pipeline replicating InstructGPT, and a robust system that optimizes training and inference, DeepSpeed-Chat claims to offer efficient, cost-effective training for models with billions of parameters. Gain broader access to advanced RLHF training with DeepSpeed-Chat, fostering innovation in AI, even for data scientists with limited resources.

Thanks for reading. FYI … I do at times use GPT-3.5 to summarize articles. I do so less to have someone/something else do the writing. It’s more to check myself and determine whether I’ve identified the important points. I hope that it improves the quality of these posts. – Rich

[nuadox]

Study: Organoids could greatly improve research on respiratory infections

- By Nuadox Crew -

Researchers at EPFL have made a significant breakthrough in bioengineering by developing organoids called AirGels, which model the human respiratory tract.

These AirGels allow scientists to study the mechanism by which antibiotic-resistant pathogens, like Pseudomonas aeruginosa, infect the respiratory tract. Biofilms, which are communities of bacteria highly resistant to treatment, have posed challenges in infection research. While biofilm formation has been studied in laboratory conditions, understanding their development in the complex environment of the human respiratory tract has been elusive.

Led by Alexandre Persat, the team at EPFL successfully created AirGels, miniature 3D tissues grown from stem cells that accurately mimic human lung tissue. The AirGels revolutionize infection research by replicating the physiological properties of the airway mucosa, including mucus secretion and ciliary beating. This technology allows scientists to study airway infections in a more realistic and comprehensive manner, bridging the gap between laboratory studies and clinical observations.

In their study published in PLoS Biology, the researchers used AirGels to investigate the role of mucus in biofilm formation by Pseudomonas aeruginosa. By infecting the AirGels with this bacterium and using high-resolution live microscopy, they discovered that P. aeruginosa induces contraction of the host's mucus using retractile filaments called type IV pili (T4P). The T4P filaments generate the necessary forces to contract the airway's mucus, allowing P. aeruginosa cells to aggregate and form a biofilm. This mechanism was validated through simulations and biophysical experiments on selected P. aeruginosa mutants.

The study demonstrates that the AirGel organoid model offers unique insights into the mechanical interactions between bacteria and their hosts' environments. This discovery of a previously unknown mechanism contributing to biofilm formation in the respiratory tract has promising implications for infection research and targeted treatments against antibiotic-resistant pathogens.

Overall, the ability to engineer organoids that faithfully replicate the mucosal environment presents new opportunities for exploring overlooked aspects of infections and investigating the impact of various physiological factors on infection development and progression.

--

Header image: Human airway epithelial cells following differentiation and development inside an AirGel tissue-engineered airway. Mucus is green, cilia is orange, actin is pink, and nuclei are blue. Credit: Tamara Rossy (EPFL).

Source: École polytechnique fédérale de Lausanne (EPFL)

Full study: Rossy T, Distler T, Meirelles LA, Pezoldt J, Kim J, Talà L, et al. (2023) Pseudomonas aeruginosa type IV pili actively induce mucus contraction to form biofilms in tissue-engineered human airways. PLoS Biol 21(8): e3002209. https://doi.org/10.1371/journal.pbio.3002209

Read Also

Organoids: Crying human tear glands grown in the lab

[deinheilpraktiker]

Forscher der EPFL und der UTHSC erforschen das Zusammenspiel von Genen, Geschlecht, Wachstum und Alter Wissenschaftler unter der Leitung des University of Tennessee Health Science Center (UTHSC) und der École Polytechnique Fédérale de Lausanne (EPFL) in der Schweiz erforschen das komplexe Zusammenspiel von Genen, Geschlecht, Wachstum und Alter und wie sie die Variation in der Langlebigkeit beeinflussen. Ihre Ergebnisse, die in der Fachzeitschrift Science veröffentlicht werden, sind ein wichtiger Schritt zum Verständnis, warum manche Menschen länger leben als andere und b... #Altern #Chromosom #Chromosom_1 #Chromosom_2 #Chromosom_3 #Chromosom_4 #Chromosom_5 #Chromosom_6 #Chromosom_7 #Chromosom_8 #Chromosom_9 #Chromosom_X #Chromosom_Y #DNA #essen #Forschung #Gen #gene #Genetik #Genetisch #Genomik #Labor #Medizin #Medizinische_Forschung #Physiologie

[transportemx]

Estudia como convertir a las carreras de autos mas sustentables

En la universidad de investigación École Polytechnique Fédérale de Lausanne (EPFL) de Suiza, un grupo de estudiantes cree que las carreras de autos pueden ser una fuerza para el bien. Tienen la misión de desarrollar autos de carreras de alto rendimiento que también establezcan un punto de referencia en la industria para la sustentabilidad. Así como la Fórmula E y la Fórmula 1, son un banco de…

[craigbrownphd]

AI Accurately Converts Brain Signals Into Video

In a ground-breaking research study published by Nature, artificial intelligence (AI) can now translate brain signals into video with significant accuracy. The AI model, named CEBRA, was developed by scientists from École Polytechnique Fédérale de Lausanne (EPFL). The possibilities of this discovery are enormous and could revolutionize our understanding of the brain using AI technology. […] The post AI Accurately Converts Brain Signals Into Video appeared first on Analytics Vidhya. https://www.analyticsvidhya.com/blog/2023/05/cebra-ai-accurately-guesses-video-watched-by-subject-from-brain-signals/?utm_source=dlvr.it&utm_medium=tumblr

[moonwatchuniverse]

45 years ago… first Swiss astronaut Claude Nicollier In 1966 Claude Nicollier became a pilot in the Swiss Air Force, switching as an airline pilot with Swiss Air by 1974. In 1978, combining this with studies in Astrophysics, Nicollier was selected a member of the first group of European Space Agency astronauts. Between 1992 & 1999 the first Swiss astronaut flew 4 space shuttle missions and maintained his role as lead ESA astronaut up to 2007. February 1996, during his third mission, STS-75, Nicollier wore the Omega Speedmaster X-33 analog-digital chronograph prototype, which was officially launched as an operational astronaut/cosmonaut toolwatch by March 1998. These official 1999 ESA/NASA portraits show him wearing a Breitling Aerospace chronograph but in May 2005, Nicollier became member of the Board of Directors of the Swatch Group and full-time Professor at the École Polytechnique Fédérale de Lausanne - Vaud CH. (Photo: ESA/NASA)