Laith Raed

I fully recommend these free hybrid workshops to everyone who wants to master the practice of deep learning, they are as practical and as interactive as it gets. The instructors are very competent and patient and don't assume any previous knowledge, except python. They are open to questions and try not to leave anything unexplained. The lessons are well prepared, supported by slides and code on publicly accessible github repos, but the interactivity and the live coding aspects make them stand out over other, some of them quite expensive practical deep learning courses.

9

1 Kommentar

I'm excited to announce that I will be giving a talk at OpenRheinMain on "Automated Policy Generation for Amazon Web Services Using Terraform-Defined Infrastructure". I will be discussing how Infrastructure as Code can be used to generate access control policies for AWS, and how to minimize their privileges during that process. 📅 When: 16:00 - 16:30 📍 Where: Room C20/3, Darmstadt University of Applied Sciences #OpenRheinMain #AWS #Terraform

13

2 Kommentare

#EurekaLabs ist ein Mensch/KI-Cyborg-Bildungs-Startup von Andrej Karpathy, Miterfinder von GPT, und ein sehr kluger und sympathischer KI-Erklärer & -Didaktiker via YouTube und Blog. Wenn ich sowas jemandem zutraue, dann Karpathy. "The teacher still designs the course materials, but they are supported, leveraged and scaled with an AI Teaching Assistant who is optimized to help guide the students through them. This Teacher + AI symbiosis could run an entire curriculum of courses on a common platform. If we are successful, it will be easy for anyone to learn anything, expanding education in both reach (a large number of people learning something) and extent (any one person learning a large amount of subjects, beyond what may be possible today unassisted). Our first product will be the world's obviously best AI course, LLM101n. This is an undergraduate-level class that guides the student through training their own AI, very similar to a smaller version of the AI Teaching Assistant itself. The course materials will be available online, but we also plan to run both digital and physical cohorts of people going through it together." https://eurekalabs.ai/

2

Is AI training a copyright minefield? ⚠️ A new study suggests it might be. „1. The exception for text and data mining should not apply to generative AI training because the technologies differ fundamentally - one processes semantic information only, while the other also extracts syntactic information. 2. There is no suitable copyright exception or limitation to justify the massive infringements occurring during the training of generative AI. This concerns the copying of protected works during data collection, the full or partial replication inside the AI model, and the reproduction of works from the training data initiated by the end-users of AI systems like ChatGPT. 3. Even if AI training occurs outside Europe, developers cannot fully avoid European copyright laws. If works are replicated inside an AI model, making the model available in Europe could infringe the “right of making available“ under Article 3 of the InfoSoc Directive. Accordingly, offering AI services to European users ultimately subjects developers to European copyright laws and European courts’ jurisdiction.“ #ArtificialIntelligence #generativeAI #intellectualproperty #copyright #textanddatamining #ChatGPT #StableDiffusion #DSMDirective https://lnkd.in/edHD72UN

1

𝐂𝐡𝐚𝐭𝐆𝐏𝐓 - 𝐈𝐦𝐢𝐭𝐚𝐭𝐢𝐧𝐠 𝐨𝐮𝐫 𝐨𝐰𝐧 𝐯𝐨𝐢𝐜𝐞: 𝐂𝐫𝐞𝐞𝐩𝐲 𝐚𝐧𝐝/𝐨𝐫 𝐚𝐧 𝐚𝐦𝐚𝐳𝐢𝐧𝐠 𝐭𝐞𝐜𝐡𝐧𝐨𝐥𝐨𝐠𝐢𝐜𝐚𝐥 𝐚𝐝𝐯𝐚𝐧𝐜𝐞𝐦𝐞𝐧𝐭? The idea that ChatGPT might suddenly start imitating your own voice during a conversation is unsettling. BuzzFeed data expert Max Woolf aptly wrote on X, "OpenAI just revealed the plot for the next season of Black Mirror." Perspective 1: The technology behind GPT-4o is groundbreaking – it can mimic human voices with uncanny accuracy, elevating AI interaction to a new level. OpenAI emphasizes that safeguards have been implemented to prevent misuse. Perspective 2: However, the fact that the AI "unexpectedly" imitated the user's voice raises serious concerns. Is it really acceptable for an AI to so easily take over your voice? And once again, a report that mainly highlights the risks. How can we arrive at a neutral assessment? 𝐖𝐡𝐚𝐭 𝐝𝐨 𝐲𝐨𝐮 𝐭𝐡𝐢𝐧𝐤? 𝐂𝐫𝐞𝐞𝐩𝐲 𝐚𝐧𝐝/𝐨𝐫 𝐚𝐧 𝐚𝐦𝐚𝐳𝐢𝐧𝐠 𝐭𝐞𝐜𝐡𝐧𝐨𝐥𝐨𝐠𝐢𝐜𝐚𝐥 𝐚𝐝𝐯𝐚𝐧𝐜𝐞𝐦𝐞𝐧𝐭? https://lnkd.in/eQBT68yb

6

3 Kommentare

💡🚀... Imagine revolutionizing your classroom with cutting-edge technology that not only sparks creativity but also transforms the entire innovation process. That's exactly what we achieved this week by integrating Large Language Models (LLMs) like ChatGPT into our "Innovation and Product Management" seminar. After the great success of our first exploration of ChatGPT in our technology and entrepreneurship seminar with Isabell Lippert (Wirtschaftsinformatik | TU Dresden) in the winter term, this week, I took a further significant step forward with 100 students by integrating LLMs into the innovation process during our "Innovation and Product Management" seminar at Technische Universität Dresden. We started by applying various creativity methods without technical assistance and then explored step-by-step how LLMs can enhance our innovation processes. 🧠✨Below, for example, you can see how ChatGPT created an image for a HORNBACH Baumarkt AG persona. 🔍 Insight 1: While it's easy to ask ChatGPT to create a persona for a specific target group, the real learning happens when students first acquire expert and methodological knowledge and then extend their solutions with LLMs. This foundation allows them to competently evaluate the AI-generated output. 📝 Insight 2: Over the years, I've noticed that students often struggle to describe a concrete persona and precisely articulate customer needs and problems. LLMs have been a game-changer in this area, broadening the horizons of everyone involved in the innovation process. Hence, I agree with Sebastian Bouschery, Vera Blazevic and Frank Piller (2023, https://lnkd.in/edTzA8jy) that LLMs contribute to broader problem and solution spaces increasing innovation performance. Thus, understanding customer needs and problems is the key for developing innovative products and services. 🤔 Insight 3: Surprisingly, I overestimated how extensively students use ChatGPT in their everyday university life and how clever they are at it. Is this good or bad? From my perspective, it's a fantastic opportunity for us educators. We can guide students through this exploration process, highlight potential pitfalls, and teach essential skills for the competent use of LLMs. 🔔 Stay tuned and follow me if you want to learn how to successfully leverage LLMs in teaching and innovation management! 🌟📚 #Innovation #ProductManagement #ChatGPT #LLMs #TeachingInnovation #CustomerNeeds #Education

26

2 Kommentare

🚀 Our latest publication on GenAI Multi-Agent Systems (MAS) 🤖👾🦾🤖 is now live on the Alexander Thamm [at] blog! 🚀 – with Philipp Schwartenbeck and Dr. Johannes Nagele Simple LLM chatbots can only take you so far. MAS represents the next evolutionary step for solving complex tasks by integrating multiple tools to revolutionize the AI application landscape. 🔍 In this article, you will get an overview of GenAI Multi-Agent Systems, including: 🔑 Key Features 🏆 Benefits 🚧 Challenges 📈 The Future Whether you're an AI enthusiast, a researcher, or a professional in the tech industry, this article offers valuable insights into the cutting-edge developments in AI technology. Don't miss out on understanding how these systems are paving the way for the future of intelligent automation.   👀📡 Stay tuned for more articles on different MAS frameworks, hardware, socio-economic implications, AI alignment, and ethics. Curious about kicking off agentic workflows in your organisation? Contact us for strategic concepts and concrete implementation solutions gained from our day-to-day business expertise. Full article:  https://lnkd.in/dU3eRxMk #GenAI #MultiAgentSystems #AI #MachineLearning #ArtificialIntelligence #TechInnovation #AIResearch #FutureTech #AIApplications #IntelligentAutomation #AIAlignment #EthicalAI

22

Hannah Blocher and I had a great time presenting our work on partial rankings of optimizers at the 12th International Conference on Learning Representations (ICLR) in Vienna. After an inspiring keynote by Mortiz Hardt (mrtz.org) on the "emerging science of benchmarks", a lot of people showed up at our poster, eager to learn how to benchmark optimizers according to multiple criteria by partial rankings. The idea is super simple: Imagine you want to benchmark #deeplearning optimizers such as adam, momentum etc... When you rank them with respect to more than one criterion, e.g., minimal loss achieved and speed, some will become incomparable. For instance, optimizer 1 might achieve lower loss at the cost of being slower than optimizer 2. Our method allows to summarize such benchmarking results in a statistically valid way without enforcing a total ranking (where all optimizers would be comparable to each other). Want to learn how this works? Check out our paper. It's a quick read of no more than two pages ;) Tiny Paper: https://lnkd.in/d-mDg4cq Code: https://lnkd.in/dFCW8tCw Poster: https://lnkd.in/dEWngSbK

47

2 Kommentare

I am pleased to announce that my recent paper "Systematic Evaluation of Online Speaker Diarization Systems Regarding their Latency" is now available : https://lnkd.in/dNwmJFeA As part of this research work, I had the opportunity to compare various open source Online Speaker Diarization systems in terms of their latency. I found it particularly interesting that end-to-end systems such as FS-EEND perform comparatively well in terms of accuracy and latency. However, the accuracy suffers greatly when the number of speakers at inference is higher than the number of speakers during training. If this problem is solved, I believe that the future of Online Speaker Diarization belongs to end-to-end systems! Currently, however, a modular system such as that provided by the DIART framework is still preferable. I would especially like to thank Prof. Dr. Sigurd Schacht and Alexander Piazza for their supportive supervision! #SpeechTechnology #MachineLearning #AI #Research

13

𝐑𝐞𝐠𝐢𝐬𝐭𝐞𝐫 𝐧𝐨𝐰 ! On 27.09.2024 we are organising a hackathon in the rooms of the Department of Computer Science at the FH Dortmund. The hackathon is aimed at students and fresh alumni from and around Informatics. On this day you will have the chance to talk to our employees from Simon Werner's team and discuss current technologies and their work with them. We look forward to meeting you! #kostal #hackathon #fhdortmund #embeddeddevelopment

52

Let me share the story behind our (HM) awarded ACM CHI Conference'24 paper “Supporting Task Switching with Reinforcement Learning”, which I consider highly exciting and satisfied what I initially expected from doing research – embarking on a long-term quest with uncertain outcomes. During my PhD, we confirmed existing findings on task-switching in the domain of automated driving. There exist better/worse timings to interrupt users and make them switch tasks, and good timings can mitigate the negative effects of uncontrolled switching. Attention Management Systems (AMS) are an old promise but are hard to build in real-time scenarios. We wanted to go beyond scripted lab studies and build a "real" AMS. But how? As we did not have the resources to create big datasets, we thought of RL as a potential solution. An RL agent could observe the user interacting with a system to deliver better timings. The agent controls the timing and is rewarded with the performance of the human interaction. If timings are bad, human reactions are too. If the timings are good, reactions may be too. Thus, we devised a dual-task balancing game, where users have to balance two balls on two platforms. Players can only operate one platform at any time, requiring them to switch between them. Since the task becomes more difficult over time, we have a clear indication of performance. We hypothesized that we could build an agent who could switch tasks so that humans play better. Still, one problem remains: how can we train such an agent? It would require endless time for humans to play so that the agent can build up a policy based on humans’ reactions. We were lucky to meet Antti Oulasvirta and Jussi P. P. Jokinen in 2022. They worked on computational rationality (CR), and it turned out that this was indeed one of the missing key ingredients for our work. Together we developed a user model mimicking human behavior to train our AMS agents. Based on the results, we conducted a user study and could confirm our hypothesis – with our agent, players could increase their performance in the balancing game, making our concept one of the first RL-based real-time capable systems to effectively support users in multitasking. We believe that our approach can be utilized in a wide range of scenarios where communication timing is essential, and we hope that the community helps in achieving this. All our software, models, etc. are publicly available. Download the tool, improve it, or use it to address other multitasking and task-switching scenarios. If you are at CHI’24, visit our presentation in the Attention and Interruptions track on Tuesday, May 14th, at 9 am, in room 313B. If not, watch our presentation video and remember – the time for attention management is now! thanks to everyone involved, i.e., Alexander Lingler, Dinara Talypova, Jussi P.P. Jokinen, and Antti Oulasvirta https://lnkd.in/dumAHDzh https://lnkd.in/d2mrRAhT

72

6 Kommentare

Agile is more than just #Scrum and #Kanban. #Agile is a mindset. Get ready now with #ICAgile.

4

MeetTheMasterClass 👨‍🎓 ✍️ 👩‍🎓 Meet one of our fellow members in TUM Entrepreneurial Masterclass Timon Grosch 👨‍🎓 💡My idea is an innovative sensor for any standard toilet that analyzes urine for a variety of biomarkers. This sensor delivers personalized dietary and lifestyle recommendations based on the analysis results. By integrating this advanced health monitoring technology into daily routines, users can proactively manage their health, detect potential issues early, and make informed decisions to enhance their overall well-being. Ultimately, this innovation aims to decrease healthcare costs by promoting preventive medicine and reducing the need for more intensive treatments. Sign up for the waitlist at vitalion-health.com 👉 What are you concretely working on within the Entrepreneurial Masterclass? Within the Entrepreneurial Masterclass, I am developing a hardware prototype for a toilet sensor as a Proof-of-Concept for my master thesis. This involves designing and integrating various sensor concepts to analyze urine for biomarkers. 👉 Why are you a part of the Entrepreneurial Masterclass? The EMC grants me access to the extensive ecosystem of UnternehmerTUM and connects me with a network of like-minded people working on entrepreneurial projects

74

Der Wettlauf geht weiter. OpenAI schickt das neue Modell GPT-o1 ins Rennen. "Similar to how a human may think for a long time before responding to a difficult question, o1 uses a chain of thought when attempting to solve a problem. Through reinforcement learning, o1 learns to hone its chain of thought and refine the strategies it uses. It learns to recognize and correct its mistakes. It learns to break down tricky steps into simpler ones. It learns to try a different approach when the current one isn’t working. This process dramatically improves the model’s ability to reason. To illustrate this leap forward, we showcase the chain of thought from o1-preview on several difficult problems below." https://lnkd.in/eXSpkG4f

3

4 Kommentare

🦸🏾💻 Application now open for 𝐇𝐞𝐫𝐇𝐚𝐜𝐤𝐚𝐭𝐡𝐨𝐧 2024, 𝐭𝐡𝐞 𝐥𝐚𝐫𝐠𝐞𝐬𝐭 𝐚𝐥𝐥-𝐟𝐞𝐦𝐚𝐥𝐞 𝐡𝐚𝐜𝐤𝐚𝐭𝐡𝐨𝐧 𝐢𝐧 𝐆𝐞𝐫𝐦𝐚𝐧𝐲 Last year I joined the HerHackathon in Mannheim. Besides working with a great team on one idea for three days, I won new connections and #superwomen supporting also my baby TechBloomz. What I valued the most was the diversity across all teams, giving some salt to all developed solutions. You can only win in participating. Very recommended! 👉 Link to application in comments #womenintech #hackathon #diversity #femaleempowerment

15

1 Kommentar

There is a way to take out more than 30% of the excess carbon globally. We are losing biodiversity at a rapid pace. Bio-diversity and carbon storage capacity in ecosystems are intrinsically linked. This means we need to solve two crises at once. The German startup biometrio.earth from Saarland measures bio diversity and therefore measures the carbon storage capacity of an ecosystem. There are so-called "empty forests" where hunters have been emptying them of fauna. These forests can't exist without fauna e.g. because of seed distribution. Therefore we have the risk that many of these forests will die and release even more carbon. biometrio.earth incentivizes a nature-based solution for carbon capture and storage. In detail, they support land owners and local communities to achieve income from alternative land use. It means the landowners don't have to slash the forest anymore to get income. They can get their income from certificates that validate that they use their land for a global benefit. However, this needs to be checked regularly to avoid abuse. Therefore, biometrio.earth monitors the fauna using wildlife cameras. What you’re doing is significant. I'm happy to have such companies in our German startup scene and looking forward to possible collaborations between you and Jina AI in the future. Digital Hub Initiative Startup-Verband #startupsummit2024

22

3 Kommentare

This one goes out to the Esports Community: As a member of the esports community (no matter whether player, researcher, lecturer, coach, manager, spectator, etc.), your insights and experiences are incredibly valuable to us. We as a research team from the University of Siegen/Germany (www.pmg.uni-siegen.de) are conducting an empirical study to explore the reasons behind the inclusivity of people with disabilities in esports. Your participation in this research will provide crucial data to better understand the current state and the motivations for inclusivity within this rapidly growing field. We have designed a brief online questionnaire that aims to assess how and why people with disabilities are included in esports and where a transfer to human resource management education could make sense. The survey will take no more than 5 minutes of your time, and your responses will be anonymous and used solely for research purposes. You can access the questionnaire via the following link: https://lnkd.in/eBfyF6Qq Additionally, we would appreciate it if you could share the link within your local esports community to help us gather a wide range of perspectives. Thank you in advance for your participation and support. Your contribution is invaluable to our research! Tobias Scholz Esports Research Network Johanna Höft

32

8 Kommentare

"A platform for platform research": Last week, our university's DFG Research Training Group Digital Platform Ecosystems (DPE) brought together researchers from all over the world in Passau to discuss one topic: the power of digital platforms. In the aftermovie, DPE spokespeople Professor Jan Krämer and Professor Andreas König explain what the #DPEForum is all about and why they were delighted to welcome the following two distinguished keynote speakers: 🎙 Prof. Dr. Martin Selmayr analysed the regulatory principles behind the EU's digital acts. We asked him whether the #AIAct could have a #BrusselsEffect similar to the #GDPR. The former Secretary General of the EU Commission is an alumnus of our university, with extensive experience in EU politics. We also asked him about a possible Passau effect. 🎙 The second keynote speaker, Professor Marshall Van Alstyne from Boston University, talked about how to combat #misinformation without regulation. As one of the founders of platform economy theory, he was something of a star in his field for the researchers at the conference. Speaking of our young researchers: We asked three junior scholars to give pitch presentations on their respective research topics. Thank you, Anastasiya Shylina, Nikolai Sachs and Tevin Tafese for taking up the challenge! 🙏 More information on the conference: 📝 Report in our University’s Digital Research Magazine: https://lnkd.in/daNAkJaB 📹  Watch the full keynote speeches on our Youtube Channel: https://lnkd.in/dgvD3V39 #DPEForum2024 #DigitalPlatformEcosystems #Innovation #Research #Collaboration #UniversityOfPassau #DFG #InterdisciplinaryConference #Passau #DigitalTransformation Deutsche Forschungsgemeinschaft (DFG) - German Research Foundation

64

2 Kommentare