Regina Jensdottir’s Post

Are you at #CPDP today? Then join us at the EDPS - European Data Protection Supervisor panel on 👉“Personal data in times of AI” (Grand Hall, 14:15)👈 Frederick Richter, Mireille Hildebrandt, FBA , Patrick Breyer MEP and Clarisse Girot , OECD will be diving into the practical dilemmas that emerge, in the context of AI, moderated by Fanny Coudert ! See you there! 👍🇪🇺 https://lnkd.in/ezzzgJHS

A new paper titled "Self-Play Fine-Tuning Converts Weak Language Models to Strong Language Models" introduces SPIN (Self-Play fIne-tuNing), a novel approach that allows LLMs to learn and grow when we Imagine an LLM playing a game against itself, constantly striving to improve.That's the essence of SPIN. The LLM generates its own training data, then tries to distinguish its self-generated responses from those created with human input. This iterative process refines the LLM's ability to understand and respond to prompts, effectively bootstrapping itself to higher levels of performance. Annotating data for LLMs is expensive and time-consuming. SPIN offers a path towards developing strong LLMs without the need for vast amounts of labeled data. By playing against itself, the LLM can continuously learn and adapt, potentially surpassing the capabilities of models trained with static datasets. SPIN demonstrates a way to fully leverage the power of human-demonstration data used in Supervised Fine-Tuning (SFT), pushing the boundaries of LLM capabilities.

🚨 **New Research from MIRAI!** 🚨 Our Research Team is thrilled to announce our contribution to the **BEWARE2024 Workshop** at the **Associazione Italiana per l'Intelligenza Artificiale Conference**! 🎉 We’ll be presenting **new results** on bias detection and mitigation in ML-based classifiers. Specifically, we introduce the concept of a **Bias Amplification Chain (BAC)**—a series of steps where bias can be amplified during the design, development, and deployment phases of trained models. As a real-world application, we demonstrate the effectiveness of this concept in the **credit scoring** domain. 📊 Additionally, our contribution includes a **quantitative analysis** using the #BRIO tool to assess and address these bias amplification patterns. Looking forward to sharing insights and advancing fairness in AI together! 💡 #BEWARE2024 #AIxIA2024 #BiasMitigation #AI #EthicalAI #ML #FairnessInAI #BiasAmplificationChain #CreditScoring #AIResearch #BRIO #TransparencyInAI https://lnkd.in/eYsdvksc

Attend this free #IITPSA Gauteng Chapter presentation where you will discover the crucial role of standards in Data Center Design by exploring topics such as building resilience into data centers, elevating statistical availability, extending standards beyond the physical layer, the ethics of AI and the importance of standards to manage it. https://lnkd.in/g34acg8D Eben Owen

Have you checked out the recently published agenda for the IJIS Institute's Justice and Public Safety #ArtificialIntelligence Summit? Learn about AI's role in the Emergency Communications field from experienced leaders Stephen Devine, Scott Brillman, ENP, CEM, Jeff Wobbleton, and Timothy Kane, ENP, CMCP. Follow the link to view the full agenda and learn about the advancement of AI in all domains: https://lnkd.in/eFFAQP5h

📢 I had the privilege of attending the BDVA - Big Data Value Association Data Week 2024 event in Leuven, Belgium, organized by LEILAC and KU Leuven. Here are the highlights from the panels I chose to attend: Day 1: 1️⃣ Advancing Data Quality: discussions on ethical AI governance, data quality tools, and FAIR principles. 2️⃣ Data Value Creation: focused on creating value from data with emerging technologies. 3️⃣ Sandboxes for Innovation: highlighted the role of controlled environments in AI innovation. Day 2: 1️⃣ Health Data Meets Synthetic Data: explored synthetic data in the European Health Data Space (EHDS). 2️⃣ AI Regulation in Europe: covered the European AI strategy and the impact of the AI Act. #DataWeek2024 #BDVA #KULeuven #DataQuality #AI #Innovation #SyntheticData #EHDS #AIAct

I'm excited to announce another important event on AI in the public sector! Organised by Insights North East and DataNEt, this knowledge exchange roundtable will bring together researchers and policymakers to explore the critical steps needed to enhance data capabilities for the effective use of AI in the public sector. More details below ⬇️

The Sysdig 2024 Global Threat Report reveals a rise in LLMjacking incidents along with evolving tactics. 📈 💰 As costs surge beyond $100k daily, LLMjacking is emerging as the priciest resource-jacking threat that organizations encounter. Stay proactive — discover strategies to manage this expensive risk and safeguard your enterprise AI by exploring Sysdig TRT's new findings on LLMjacking. 👉 https://okt.to/JeWDSs

Learning from experts in the field of Trust, Security and Compliance with AI-based applications.

 Thora Markert and Vasilios Danos from TÜV Informationstechnik GmbH - TÜVIT (TÜV NORD GROUP) were sharing the expertise on the process and challenges of implementing trustworthiness in AI at this years applydata summit. This keynote explores the unique challenges and risks associated with AI-based functionality in security-critical applications and examines the impact of upcoming AI regulations on the development and deployment processes across the entire life cycle of an AI system. Check out the video below and learn more about the gaps and challenges of AI-based Application. https://lnkd.in/eKzqyjud #applydatasummit #datasummit #AI #EUAIact #Trust #TrustworthyAi #keynote

🌐AI & Partners - International Business Times UK - Navigating the Global #AI Landscape: Balancing Regulation and Innovation Exciting times ahead for the global #AI landscape! Check out this insightful article discussing the crucial balance between regulation and innovation in #AI development. 🤖💡 Regulation is essential to ensure #AI aligns with societal values and norms, mitigating risks like biased algorithms and privacy breaches. However, overly prescriptive regulations can hinder innovation, especially for start-ups and small businesses. Global collaboration is key amidst geopolitical tensions, fostering initiatives like GPAI and OECD.AI’s #AI Principles. Finding the right balance between regulation and innovation is crucial. Thanks to Informa for helping co-write this article. Their in-depth insights as well as #AI acumen really drove forward, amongst other things, the content generation process. 🚀 Join AI & Partners at The AI Summit London 2024 to delve deeper into these topics and network with industry leaders! Register now for an enriching experience in #AI advancement. Registration details: 📅 Date: June 12 – 13, 2024 📍 Location: Tobacco Dock, London 🎟️ Register here: 15% discount code: AIPARTNERS15OFF Register via: https://lnkd.in/dpFwh3zt Website link: https://lnkd.in/dp2BA7Qi Don't miss out on shaping the future of #AI! #AISummitLondon #AI #Innovation #Regulation The AI Summit Informa Emily Goracke Dee Saniukaite