Sebastian Modlich

🚀 Exciting News! 🚀 We are thrilled to announce that Kumo AI's position paper "Relational Deep Learning: Graph Representation Learning on Relational Databases," has been accepted at #ICML 2024! 🎉 Join us in Vienna from July 21-27, where our very own Matthias Fey will be presenting this exciting new research area. 🔗 https://icml.cc/ In collaboration with Stanford University, we introduce Relational Deep Learning (#RDL), a new research area for applying deep learning on relational databases by viewing them as temporal, heterogeneous graphs. This method eliminates the need for manual feature engineering, leading to more accurate models built faster. 🚀 We’ve also developed #RelBench, a benchmark suite to facilitate research in this new area, using diverse datasets from Stack Exchange discussions to Amazon book reviews. 📚 Stay tuned for more updates on our presentation format, and check out our paper here: https://lnkd.in/dk24vbSg #ICML2024 #MachineLearning #AI #GraphNeuralNetworks #RelationalDeepLearning #DataScience #KumoAI #StanfordCollaboration

38

Join us as we unpack the future of AI with Carsten Kraus in this follow-up article from Launch's #NavigatingAbroad series. Learn about the impactful role of small language models in making AI more accessible and the intricate dynamics of reinforcement learning. What implications will these technological advances have on the future landscape of AI? #AIUnpacked #FutureOfAI #LanguageModels #ReinforcementLearning #TechInnovation

6

📅 Mark your calendars for September 18th in #Vienna! At the #DataScience MeetUp, Andreas from Neo4j and Sebastian will talk about the relevance of #KnowledgeGraphs for Gen #AI, and explain why proper #GraphVisualization is critical for understanding your #data. We look forward to seeing you there! https://lnkd.in/e-EKZ9fZ

6

#EarlyCareer #talk at #IJCAI2024 by🗣️Julia Neidhart, Technische Universität Wien: Transforming Recommender Systems: Balancing Personalization, Fairness, and Human Values ➡️https://lnkd.in/dK3Nieyr Abstract: Recent advancements in recommender systems highlight metrics beyond accuracy, including diversity, serendipity and fairness. This paper examines the integration of Large Language Models into recommender systems, focusing on challenges such as interpretability, bias, and user autonomy. LLMs offer enhanced personalization capabilities but also introduce concerns regarding transparency and ethical implications. This work discusses ongoing research efforts aiming to develop transparent, fair, and contextually aware systems. Our approach seeks to prioritize user wellbeing and inclusivity, contributing to a more equitable and functional digital environment through advanced technologies and interdisciplinary insights. #XAI #LLMs #InclusiveTech #NextGenAI #AI TU Wien Informatics

3

New German/English AI Model: A Big Step for Language Tech Have you ever wondered why most language AI seems to work best in English? Well, that's changing! A new AI model called deepset-mxbai-embed-de-large-v1 is here to help with German and English languages. Let's break down what this means and why it's exciting. What Is This New Model? The deepset-mxbai-embed-de-large-v1 is a tool that helps computers understand German and English text better. It's like a super-smart language helper for machines. Two companies, Deepset and Mixedbread, worked together to create it. Why Is It Special? This new model is really good at understanding context in German and English. It can help find the right information quickly, which is great for things like search engines or chatbots. The cool part? It's free for anyone to use and improve! How Does It Work? The team trained this model using millions of German sentences. They taught it to understand the meaning behind words, not just the words themselves. It's like teaching a computer to read between the lines. What Can It Do? This model can help with lots of language tasks: -Finding the right information in a big pile of text -Sorting documents by topic -Understanding if a sentence is happy, sad, or angry -Helping translate between languages Why Should We Care? Most language AI tools work best in English. This new model helps level the playing field for German speakers. It means better tech tools for German-speaking countries and anyone working with both German and English. How Good Is It? The team tested their model against others, and it did really well. It scored higher than older models in understanding and sorting information accurately. This means it's not just new, but also better at its job. Easy to Use The best part? Developers can easily add this model to their existing projects. It's like plugging in a new, more powerful engine to make your car run better. This means we might see improvements in many apps and websites soon. What's Next? This model is just the beginning. As more people use and improve it, we could see even better language tools in the future. It might lead to smarter chatbots, more accurate translation apps, or even help computers write in German more naturally. Why It Matters In today's connected world, it's important that technology works well in many languages, not just English. This new model is a big step towards making AI more inclusive and helpful for German speakers. Model - https://lnkd.in/g3pk5kKE Follow AI Toolhouse - AI Tools Catalogue for more insights and visit AI Toolhouse for 3700+ productivity tools! 🔗https://lnkd.in/dpQB7xZU ---- 🤖 Contact us if you made a great AI tool to be featured - https://lnkd.in/d_HbS8D3 #GermanNLP #AILanguageModel #DeepLearning #NLP #MultilingualAI #OpenSourceAI #GermanAI #LanguageTechnology #MachineLearning #AIInnovation #AIResearch #LanguageAI #TechInnovation #GermanTech #AIAdvancement #DataScience

3

RAG Sytems are not just about using powerful LLMs. What do you do if pre-trained embedding models are not working well for retrieval? Full example from dataset creation to fine-tuning in Module 4: https://lnkd.in/e8nWipvd

5

The Rails 8 feature set perfectly complements SQLite's power in creating resilient, high-performance production apps, but still the question lingers: Can I really go all-in on #SQLite? Join Stephen Margheim at #RailsWorld as he illustrates how to leverage Rails and SQLite's full potential in your next venture, and when SQLite does and doesn't make sense for your application. https://lnkd.in/eaBstScy

52

1 Kommentar

I'll never be able to read the word "delve" again without immediately thinking: Ah, this is AI writing. Anyway! A new paper has been published, along with fine-tuned models for generating really long outputs. The researchers weren't satisfied with the fact that current LLMs could technically produce much more output than they normally do. ChatGPT and similar models typically generate around 2000 words, despite being capable of producing far more tokens. Their hypothesis: The models don't encounter longer outputs during training. Initially, the researchers used prompt chaining to make ChatGPT-4o generate texts longer than usual. After succeeding, they created a training dataset with longer texts and fine-tuned two current models, Llama 3 and GLM 4. I tried the GLM 4 on Hugging Face and fed it a brief book summary of Jonathan Safran Foer's "Eating Animals." The result is ten pages of text, approximately 30,000 characters. The, well, good news: The result isn't a copy of Foer. It's AI slop, barely readable at best. Which is also the bad news, I guess.

30

7 Kommentare

📅 On 07.11.2024, we invite you to the ScaDS.AI Dresden/Leipzig training "Big Data Processing on HPC" with Apurv Kulkarni and Wenyu Zhang. Apache Spark and Apache Flink are two typical #BigData analytics frameworks. Their APIs allow the development and testing of an application on a local workstation and later, without changing the source code of the application, distribute work to many computers when the local workstation is not sufficient anymore due to limited resources. The course focuses on the step from a local workstation to an #HPC environment and presents how the typical Big Data analysis workflow can be organized in an HPC environment. In this course participants will be introduced to running a data pipeline and data processing along with managing the configurations on the HPC environment, using Apache Flink and Apache Spark. Register here: 👉 https://lnkd.in/eYfmiMT8

15

For all the wizardly power of Large Language Models (LLMs), they cannot reliably handle math. In our world of strategic insights, we need both deep conversations and rigorous statistics, integrated into one cohesive story. Tools like ChatGPT and Claude are game changers when it comes to analyzing the "conversation" part of the equation. To overcome their inherent weaknesses in reliably computing math and statistics, we utilize something they are now very good at, Python programming. By leveraging Python's robust libraries, LLMs can bypass their mathematical constraints and execute precise computations. Whether you're a data scientist, financial analyst, or researcher, you can learn how integrating LLMs with Python can enhance your analytical tasks, from simple calculations to complex forecasting models. Our latest piece shows how allowing LLMs to utilize Python programming democratizes access to advanced computational tools, enabling users with limited programming expertise to find insights from sophisticated mathematical and statistical techniques. #ArtificialIntelligence #MachineLearning #DataScience #PythonProgramming #LLMs #MathInAI #DataAnalysis https://lnkd.in/gq-QWqig

7

🔎 𝗫𝗔𝗜 𝗖𝗼𝗻𝗳𝗲𝗿𝗲𝗻𝗰𝗲 𝟮𝟬𝟮𝟱 - 𝗦𝗽𝗲𝗰𝗶𝗮𝗹 𝗧𝗿𝗮𝗰𝗸 𝗦𝗽𝗼𝘁𝗹𝗶𝗴𝗵𝘁: XAI Engineering 🖥️ ✏️ 𝗔𝘂𝘁𝗵𝗼𝗿𝘀: Severin Kacianka Verena Klös Astrid Rakow Maike Schwammberger 🔗 𝗟𝗲𝗮𝗿𝗻 𝗠𝗼𝗿𝗲 𝗛𝗲𝗿𝗲: https://lnkd.in/dUYbTDkV Explainability is the requirement for AI systems to provide human-understandable reasons for the output and actions of the system. For this, it must be taken into account that such AI systems are generally comprised of three kinds of components: 1) Symbolic AI, also known as classical AI, uses approaches of logic, programming and classical engineering, 2) sub-symbolic AI uses approaches of statistical learning, often referred to as machine-learning, and, 3) traditional non-AI software using deterministic programming. Explainability of such AI systems can target models and algorithms at design time and context-dependent system behaviour at run-time. In the latter case, we call a system self-explainable if it can autonomously decide when, how and to whom it should explain a topic. To develop such self-explainable systems, explainability needs to be carefully considered during the whole systems engineering process: discovering requirements related to explanations (e.g., comprehension problems, timing, interaction capabilities, different needs of stakeholders), design methodologies, tools and algorithms for (self-) explaining systems, as well as testing explainability. Currently, there is only little research on systematic explainability engineering; instead, we face a long list of open research questions in different system engineering disciplines. These challenges must be tackled on an interdisciplinary level, integrating expertise from various areas of computer science (e.g., systems engineering, AI, requirements engineering, HCI, formal methods, logic) and social sciences (e.g., law, psychology, philosophy). This track tackles systems engineering perspectives on explainability, intending to join forces to make complex systems of systems with AI components explainable. #explainable #artificialIntelligence #system #engineering #symbolicAI #subsymbolicAI #machinelearning #deterministic #programming #selfexplainability #systematic #explainability

11

Today's reading is Memory Layers in Large Language Models 🧠 This research shows how memory layers can take LLM performance to the next level: 🔑 Trainable key-value lookup mechanism: Adds parameters without increasing computational cost, boosting factual accuracy and task performance. 📈 Big results, small cost: Outperforms models with much larger computational budgets and even beats mixture-of-experts models! ⚙️ Scales seamlessly with up to 128 billion memory parameters The authors from Meta highlights how memory layers are a game-changing addition to future AI architectures, offering smarter, faster, and more accurate solutions. You may want to check out the paper at: https://lnkd.in/gzuqUfYe #AI #LLM #MemoryLayers #Innovation #MachineLearning #FactualAI #FutureOfAI #Meta

19

Want to excel as a machine learning engineer? In his recent video Boris Meinardus a student researcher at TU Darmstadt, summarizes the seven most important skill sets based on insights from ML engineers at top companies like Google, Meta, NVIDIA, and OpenAI. Key Skills: 1. Programming Skill 2. Mathematics 3. Soft Skills 4. Familiarity with Core Libraries 5. Machine Learning Knowledge 6. Model Evaluation 7. Software Engineering Skills

1

The gap to existing state-of-the-art LLMs Extended Long Short-Term Memory (xLSTM), a type of neural network model designed to process sequences of data, such as in language modeling. At the ICLR 2024, Sepp Hochreiter and team unveiled the xLSTM, a new AI architecture. It introduces new features like exponential gating and updated memory structures to improve performance over traditional LSTMs and compete with newer models like Transformers. The xLSTM achieves better results by enabling parallel processing and using advanced memory techniques. This makes it more efficient, particularly for tasks involving long sequences of data.

9

4 Kommentare

Let's Understand & Build an #AgenticRAG Pipeline in 13 Mins!

2

Using LLM to create ontologies from knowledge that the LLMs have been trained on CSM provide a core ontology seed that can be extended for a domain. GPT4 seems to have capability with good promos to be able to express an ontology in Tirtle format and you can even set the uri and a prefix. Choose some domain that is fairly stable in concepts and start to question about core concepts on this domain eith an express goal for the use of the ontology. Then ask about the relationships. Use these results on another prompt to crest an OWL ontology in Tirtle format with the uri you define snd use the 3 letter prefix. Then ssk to populate it eith test data, then ask to crest SPARQL queries. Once you have through prompts validated your ontology the ssk it to create new classes with existencecacioms to infer new classification of instances that satisfy the requirements for membership you state in natural language. Again ask it to test with data and ask for the inference results. Validate by copying tirtlecdyntsx ontology to Protege and validate. I have use this process also with cternsl documents to create ontologies representing the concepts relationships for documents having the goal of defining compliance, and other models of concepts for understanding some domain such as policies.

11

2 Kommentare

Deep Dive into xLSTM with Max Beck at ML Week Europe Join us at Machine Learning Week Europe for a keynote by Maximilian Beck from Johannes Kepler Universität Linz and our NXAI team. Max will talk about “xLSTM: New Architectures for Large Language Models”, sharing his latest breakthroughs in efficient and scalable AI models. 👉 Ready to explore the future of AI? Join Max and the NXAI team in Munich at Machine Learning Week! #MachineLearning #xLSTM #AI #MLWeekEurope #NXAI Peter Seeberg, Martin Szugat

111

4 Kommentare