Ludvig Strand

Finnish researchers have developed a technology that enables computers to understand human emotions to improve AI interactions, potentially allowing machines to adjust their behavior to enhance user experiences. Researchers at the University of Jyväskylä in Finland have created a model that allows computers to recognize and comprehend human emotions by employing mathematical psychology principles. This breakthrough could enhance the interaction between humans and smart technologies, such as artificial intelligence systems, by making them more attuned and reactive to users’ emotions. In everyday interactions with computers, users commonly experience emotions such as joy, irritation, and boredom. Despite the growing prevalence of artificial intelligence, current technologies often fail to acknowledge these user emotions. The model developed in Jyväskylä can currently predict if the user has feelings of happiness, boredom, irritation, rage, despair, and anxiety... #artificialintelligence #technology #machinelearning #breakthrough #innovation #computers #emotions #humans #research #psychology

27

14 kommentarer

We’re honoured to welcome Matthias Seeger, Principal Applied Scientist at Amazon AI, to the leading data science, ML, and AI event in the Nordics - NDSML Summit 2024, scheduled for the 23rd and 24th of October in Stockholm! On the Machine Learning + MLOps, Matthias will be presenting on "Challenges of Evaluating AI-Powered Troubleshooting Assistants". Matthias is a leading mind in machine learning and one of the scientists behind Amazon Q! With a Ph.D. from Edinburgh University, postdoc work at UC Berkeley, and a career spanning roles at the University of Saarbrueckens and EPF Lausanne, Matthias has made a huge impact in the AI and tech space. Since joining Amazon in 2014, he’s helped revolutionize how customers build, optimize, and operate applications on AWS. Oh, and did we mention he’s also an ICML Test of Time Award winner? Don’t miss this opportunity to meet Matthias and 400+ like-minded practitioners! Book your tickets at https://hubs.li/Q02RsQKs0 #NDSMLSummit #DataScience #AppliedDataScience #MachineLearning #ML #ArtificialIntelligence #AI #EnterpriseAI #DeepLearning #DataEngineering #MLEngineering #DataPipelines #BigData #Automisation #GenerativeAI #EthicalAI #TrustworthyAI #LLMs #DataOps #MLOps

10

🎙️ #TheDataDemocracyPodcast Episode 29 is out now! 𝐌𝐞𝐬𝐡𝐞𝐝 𝐔𝐩 𝐓𝐞𝐫𝐦𝐬: 𝐀𝐥𝐢𝐠𝐧𝐢𝐧𝐠 𝐰𝐢𝐭𝐡 𝐃𝐚𝐭𝐚 𝐌𝐞𝐬𝐡, 𝐃𝐞𝐛𝐚𝐭𝐢𝐧𝐠 𝐭𝐡𝐞 𝐃𝐢𝐜𝐭𝐢𝐨𝐧𝐚𝐫𝐲 ✔️📖 In this podcast episode, host Ole Olesen-Bagneux sits down with Jon Cooke, the founder of Dataception and a seasoned pioneer in the realms of data architecture and AI. He takes us through his diverse career journey, starting from his early days in virtual reality and aerospace to spearheading innovations in data strategies and AI at major consultancies and enterprises. Jon delves into the practical challenges of implementing data mesh architectures, sharing his nuanced views on the transition from vision to execution. He critiques common terminologies like 'data products’, proposing clearer communication to bridge the gap between technical teams and business stakeholders. Throughout the conversation, he emphasizes the importance of composable data architectures and the evolving role of generative AI in transforming how businesses handle and leverage unstructured data. As the discussion unfolds, Jon highlights his current work on cutting-edge projects that integrate AI with traditional data processes, reshaping risk assessment and decision-making frameworks. His insights offer a glimpse into the future of data management, where AI and data mesh principles converge to create more adaptive and intelligent enterprise ecosystems. 🎥 Watch this episode on YouTube: https://hubs.ly/Q02t6ZQd0 📻 Listen to this episode on your favorite podcast platforms: https://hubs.ly/Q02t71rb0 #podcast #datademocracy #datamesh #dataproducts #AI

12

2 kommentarer

What a great experience speaking at the Nordic Data Science and Machine Learning Summit 2024! 🙌 I had the opportunity to present on “Data Platforms - Generating Value from Analytical Data at Scale”, diving into how an opinionated data platform can help organizations build and operate data flywheels to unlock real business value. 🔄 The questions and discussions—both during the Q&A and later during the break—were insightful and showed how much we all care about tackling data engineering challenges together. 💡If you were there and didn’t get the chance to ask questions or discuss the topic, don’t hesitate to connect and DM. I also thoroughly enjoyed the session by the great storyteller Lars Albertsson on “Generative AI - the power to destroy democracy meets the security and reliability of MS-DOS”. What a thought-provoking talk! 🧠 Big thanks to the #NDSMLSummit team for putting together a well-organized event, and to everyone who attended and contributed to the lively discussions. 🙏 #DataEngineering #DataPlatforms #DataFlywheels #GenerativeAI #MachineLearning #NDSMLSummit

90

12 kommentarer

Dear all it's Friday afternoon and next week is the best week of the year: It is Data Innovation Summit in Stockholm week. As you can tell by the illustration I AM GOING! Next monday I will publish some thoughts about my talk - a highlight of the roadmap we have in Zeenea and how we can mirror the mesh. There are already many people signed up for my presentation, so I suggest you download the Agorify app, and register now, if you want to save your spot. Also come and visit Luc Legardeur Petter Tønnessen Anthony Corbeaux and me in the Zeenea booth for a talk on our ideas. I hope to see all of my friends in the Nordics next week!! #datamesh #data #datacatalog #datamarketplace #datamanagement #federation #domaindrivendesign #domain #knowledgegraph #metadata

46

2 kommentarer

🌟 Ever wondered if AI is really making us better or if it’s all in our heads? 🤔 A new study by Aalto University shows that the belief in AI's power can boost human performance due to the placebo effect! 🚀💡 Participants performed tasks more efficiently when they thought an AI was assisting them, even though no real AI was involved. 🔍✨ Expectations alone can make a significant difference! 📊 Discover how the placebo effect shapes our perceptions of AI and its implications for future evaluations. 💭📈 The study reveals the hidden power of our expectations and how they influence our interaction with technology. 🌐🔍 Join the conversation on the fascinating intersection of psychology and technology. 💬🤖 Read more about this intriguing study and share your thoughts! 🔗🗨️ https://lnkd.in/gkiMxtjB Karmactive #AI #PlaceboEffect #TechResearch #HumanPerformance #AIStudy #Psychology #TechTrends #AaltoUniversity

5

This Fall we are teaching a course on LLMs and Generative AI at the University of Helsinki, together with Aarne Talman, PhD (Accenture) and Jussi Karlgren (Silo.AI, now AMD). Course page: https://lnkd.in/dqHAhEPA Aarne and Jussi took care of the foundation: intro to LLMs, prompt engineering, fine-tuning and evaluation. I gave the lecture on RAG (Week 5), which included a live demo built with Gemma LLM, SBERT and streamlit, allowing you to query your PDF files with natural language questions. The code runs locally, which means there is no dependency on any API provider. This certainly also means, that LLM inference may be much slower if ran on CPU (though having a local GPU isn’t a rarity today!). However, such an approach has certain benefits, for example the ability to query a sensitive PDF (say, a statement from your bank) without worrying that your precious data goes onto the cloud. In the blog post you will also find the lecture recording. I’d like to thank Aarne Talman, PhD for inviting me to co-teach this course, and to Atita Arora for helping out with some RAG materials and sharing her slides. In the lecture I have also relied on the work done by Daniel Bourke.

76

8 kommentarer

Last week, I had a pleasure to teach the Week-6 topic: "Use cases and applications of LLMs". We looked at multimodal LLMs, as a very interesting, and in many ways still an emerging trend in the LLM world, covering text, image, video and audio modalities (you can ask: "What do you hear in this video?", for example, - besides a more regular "Does the man wear glasses?"). We covered PaliGemma, ColPali, Video-LLaMA models. I've also demoed various industry-grade LLM applications, including: * TomTom's Intelligent Vehicle Assistant - shoutout to Massimiliano Ungheretti, PhD (who starred in the lecture) and IVA team Ola Aleksandra Ferdzyn-Grzelak, Alexey Samuleenkov Cezary Draus and the rest of the team * BabyAGI - built by a VC, showing the power of modern LLMs to solving multihop challenges, like assembling a list of under-represented topics for a blog post * Code assist (GitHub Copilot, Cursor) and "office-assist" * Multimodal vector search we have built with Aarne Talman, PhD for the GSI Technology In the end we had a fun group brainstorming session, where students came up with the fields deserving to be improved by LLMs, benefits, risks and their mitigation. Sessions like this sparkle creativity in the audience and turn the class around (students help me complete the slides, which is always cool and appreciated). In the lab we looked at: 1) querying tables (extracted from PDF files) using natural language: "What was Google's operating margin for 2024?" -> "According to the table, Google's operating margin for 2024 is 32%." 2) generating synthetic data Link to the YT recording in the comments below. Follow the podcast for more content and interviews around LLMs, search, RAG, and emerging trends.

28

2 kommentarer

🎓 Workshop Recap: Leveraging Generative AI at SITE 🎓 Today, I had the pleasure of presenting two workshops at the Stockholm Institute of Transition Economics / Östekonomiska Institutet (SITE) at the Stockholm School of Economics. These sessions focused on how generative AI can augment and improve workflows for researchers and administrators alike. Key highlights from each session: 💼 For Administrators: • Demonstrated how AI can optimize communications, budget management, and event planning using tools like CustomGPTs and Claude Projects. • Explored practical use cases for Microsoft Copilot to accelerate workflows in document summarization, email drafting, and summarizing meeting minutes. • Provided strategies for navigating AI risks, such as data leakage and prompt injection, to ensure safe adoption. 📚 For Researchers: • Streamlined research processes through tools like ResearchRabbit and Elicit, enabling faster literature reviews and enhanced analysis. • AI-driven ethics in academia, with discussions on transparency, human contribution, and acknowledgement of AI’s role. • Prompt engineering with the SCRIBE Method to ensure outputs are accurate, reliable, and impactful. A huge thank you to Dominick N. and Minh Phan for hosting these sessions and fostering innovation at the intersection of academia and AI. It’s always inspiring to help teams unlock the potential of generative AI to make their work more efficient, ethical, and impactful! #GenerativeAI #Academia #Innovation #PromptEngineering #AIinAdministration #SSE

6

2 kommentarer

const textSession = await window.ai.createTextSession(); 🤯

5

Arguably the most important of the 8 AI ethics principles, ensure your AI system does not perpetuate bias, it must never discriminate against a particular group or demographic, and it should be fair, inclusive and accessible. Ensure your AI committee or go-no go decision makers represent a diverse group of individuals, with the fundamental tenet to do no harm. #privacy #privacymatters #AIethics

2

Simplify the evaluation & implementation of AI by being unapologetically practical & objective of how it can be beneficial to you. Read these wise words from Lasse Rindom below. My top 3 favorites are: 💡 How people and enterprises work with AI 💡 What AI can and cannot do 💡 How AI works when touching the real world #ai #genai #innovation

14

3 kommentarer

It feels strange to get excited when talking about AI and sexual and violent abuse of children. But this is an area where AI can be an incredibly powerful tool to prevent it, and these types of use cases don’t get enough attention. Today, I visited the World Childhood Foundation at their new office in Stockholm to discuss how AI can be applied in this field, together with Stiftelsen Allmänna Barnhuset. They conduct great research and projects on this topic, and AI should enable them to achieve even more. Abuse of children is a tough subject to process (especially for my 5:00 AM train-ride brain), but AI has great potential to help preventing it. These are the problems and solutions that should be at the forefront when we talk about what AI can do for us. If half the energy people spend showing GPT-4 can’t count the letter 'R' was redirected toward thinking about this, we could unlock the potential of AI for truly impactful work.

43

7 kommentarer

Good data hygiene and management is critical for both Cloud and AI transformation.

1

This is the second of eight voluntary AI Ethics Principles from CSIRO's Data61 and Department of Industry, Science and Resources, designed to guide businesses and governments to responsibly design, develop and implement AI. Every day, we showcase one of the eight AI Ethics Principles. Calabash Solutions. Working with organisations to Start. With Privacy. #privacy #privacymatters #aiethics

4

I start to believe we have passed the top of the #AIHypeCurve. Last week has been the most noisy week with claims like «Arendalsuka: AI is a revolution», «Sweden: This is how you AI-wash your campaign», «Norway must become the leader of AI in Europe», «Denmark breaks the AI code», «New competitive edge for Finland with AI».Politicians and responsible business leaders have all been jumping on the train of big claims. The trouble with all these lofty statements is that they lead to unrealistic expectations and take the focus away from the proven value we already see in areas such as IT development, operations, risk management and customer service. The businesses who succeed with AI are doing careful, planned implementations with a clear tracking of benefits which are articulated, calculated and driven by knowledgable people. There is no magic, and high expectations alone will not get us anywhere. The report GENERATIVE AI IN ORGANIZATIONS 2024 by the Capgemini research Institute is well worth the read. Published in July with real usecases and data. Just search for it online. #NordicAIFrontier #GenerativeAIInOrganisations Caroline Wilhelmsen Thordur Arnason Thomas Svahn Michael Jacobs Samay Mir Johanna Staaf Arash Afsarian Alexander Fritsch Mikko Misukka

111

11 kommentarer

Following our discussions at MeasureCamp, Steen provided an excellent summary of the challenges analysts face in communicating results. Take a look and discover 6️⃣ ways to overcome the communication gap.

6

1 kommentar

Niclas got me thinking about knowledge sharing within the analytics context. The obvious angle is to organize some form of show and tell, (e.g. daily standup) where you can show how you did something, how to use a technology, or explain conclusions. Maybe the collective memory of the team will remember when this info becomes relevant, likely not. But I wonder - how does one convey the thought process that went into investigation, pattern recognition, applying business context to the data question at hand. After all, analytics is closer to detective work than anything else. Detectives string together many bits and pieces together, both newly dug up info as well as notes from old case files. Pretty sure in analytics we do something similar. - Talk to a stakeholder, find out there's another person to talk to - Stalk how current business processes are done in real life - Remember to exclude that data anomaly from 1.5 years where revenue got counted twice - Hunt down that one query you used few months ago that could repurposed Before the summer we spoke to a ton of analysts and a good majority cited that most time went into the context and investigation process. Mangling data close second, but the actual analysis part was quote - the easiest. So, if the majority of time is spent on the "WTF" phase, and every analyst goes through a personal, daily "WTF" phase, wouldn't it make sense to focus on the skills needed to get out of WTF faster? Less mistakes, more thorough work, faster turnaround time. Some of the brightest analysts I've met have a knack, natural or trained, for this. But - seems that currently the only real way to acquire this type of knowledge is with hands-on experience and time. -- #dataanalytics #analytics

11

1 kommentar

Today Peter Danenberg and I recorded one of my favorite sessions so far. "LLMs the Language of Creativity, Power and Intelligence." We took a philosophical approach to the topic of LLMs and discussed questions such as: ✔ What is the inherent power of language? ✔ How did Large Language Models create the first thinking 'being', capable of being more intelligent than humans? ✔ Is hallucination inherently bad or is it indeed the source of creativity? ✔ Is Gen-AI creative? ✔ Could we ever truly measure the intelligence of AI models if they surpasses human intelligence? Love to hear your feedback. BTW, I looked up the highest intelligence recorded is YoungHoon Kim, south Korean citizen, with a recorded IQ of 276. With that said, the argument about measuring IQ of 200 vs IQ of 2,000 still stands.

39

10 kommentarer