<?xml version="1.0" encoding="utf-8"?><feed xmlns="http://www.w3.org/2005/Atom" ><generator uri="https://jekyllrb.com/" version="3.10.0">Jekyll</generator><link href="https://deepdivepod.eu/feed/episodes.xml" rel="self" type="application/atom+xml" /><link href="https://deepdivepod.eu/" rel="alternate" type="text/html" /><updated>2026-10-28T04:38:11+00:00</updated><id>https://deepdivepod.eu/feed/episodes.xml</id><title type="html">Deep Dives with Iman | Episodes</title><subtitle>Exploring the science, tech and societal impact of artificial intelligence with accessible expert insights.</subtitle><entry><title type="html">Teaching Robots Skills on Real Hardware — Jens Kober</title><link href="https://deepdivepod.eu/episodes/32/" rel="alternate" type="text/html" title="Teaching Robots Skills on Real Hardware — Jens Kober" /><published>2026-10-22T00:00:00+00:00</published><updated>2026-10-22T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/32</id><content type="html" xml:base="https://deepdivepod.eu/episodes/32/"><![CDATA[]]></content><author><name></name></author><category term="Technology" /><category term="Robotics" /><category term="Reinforcement Learning" /><category term="Imitation Learning" /><category term="Reward Hacking" /><category term="Simulation" /><category term="Humanoid Robots" /><category term="Uncertainty" /><category term="TU Delft" /><category term="Stuttgart" /><summary type="html"><![CDATA[Deep Dives with Iman, hosted by Iman Mossavat, features Professor Jens Kober, head of the Learning and Interactive Robots Lab at the University of Stuttgart, formerly of TU Delft. Kober has spent much of his career on one question: how do you get a robot to learn a skill and perform it reliably on real hardware? We discuss why tasks that look trivial to humans, such as picking up a mug, are so hard for robots, why flashy humanoid demo videos say little about reliability in a factory, and why learning in simulation does not simply carry over to the real world.]]></summary></entry><entry><title type="html">Moving the Eiffel Tower: Opening LLMs with Explainable AI — Jelle Zuidema</title><link href="https://deepdivepod.eu/episodes/31/" rel="alternate" type="text/html" title="Moving the Eiffel Tower: Opening LLMs with Explainable AI — Jelle Zuidema" /><published>2025-10-27T00:00:00+00:00</published><updated>2025-10-27T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/31</id><content type="html" xml:base="https://deepdivepod.eu/episodes/31/"><![CDATA[]]></content><author><name></name></author><category term="Technology" /><category term="Large Language Models" /><category term="Explainable AI" /><category term="Interpretability" /><category term="Transparency" /><category term="Reasoning" /><category term="Deep Learning" /><category term="Bias and Fairness" /><category term="Amsterdam" /><category term="Netherlands" /><summary type="html"><![CDATA[Deep Dives with Iman, hosted by Iman Mossavat and featured by Radio 4 Brainport, features Associate Professor Jelle Zuidema from the University of Amsterdam's Institute for Logic, Language, and Computation (ILLC). Zuidema analyzes large language models (LLMs) using explainable AI, including striking examples to show how model knowledge can be surgically edited. He briefly discusses how researchers understand LLMs, fix biases, and analyze how they reason. Listeners gain insight into how these models represent knowledge, interact with data, and how analysis can clarify complex AI systems. This episode gives you an excellent view into the inner workings of modern language models.]]></summary></entry><entry><title type="html">AI Reshaping Software Engineering &amp;amp; Education — Alexander Serebrenik &amp;amp; Lina Ochoa</title><link href="https://deepdivepod.eu/episodes/30/" rel="alternate" type="text/html" title="AI Reshaping Software Engineering &amp;amp; Education — Alexander Serebrenik &amp;amp; Lina Ochoa" /><published>2025-10-15T00:00:00+00:00</published><updated>2025-10-15T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/30</id><content type="html" xml:base="https://deepdivepod.eu/episodes/30/"><![CDATA[]]></content><author><name></name></author><category term="Technology" /><category term="Software Engineering" /><category term="Large Language Models" /><category term="Code Generation" /><category term="Technical Debt" /><category term="Evaluation and Testing" /><category term="Education" /><category term="Future of Work" /><category term="Human-AI Collaboration" /><category term="Open Source" /><category term="TU Eindhoven" /><category term="Eindhoven" /><summary type="html"><![CDATA[How should software engineering education adapt when LLMs can generate code in seconds? Prof. Alexander Serebrenik and Dr. Lina Ochoa of TU Eindhoven join Iman Mossavat, host of Deep Dives with Iman, to examine how large language models are reshaping software engineering and education. We cover: how models optimise for user satisfaction and why verification remains essential; the shifting skillset for engineers (from rote coding to oversight, orchestration, and prompt decomposition); real quality risks such as hallucinations and increased technical debt when generated code is used without enough scrutiny; and the structural effects as model providers gain influence. Practical conclusions: keep humans in the loop, teach verification and system thinking, and redesign assessments to preserve core understanding. Listen for a measured discussion relevant to educators, engineers, and policy makers.]]></summary></entry><entry><title type="html">How Do We Know AI Gets Human Thinking Wrong? — Benjamin Paaßen</title><link href="https://deepdivepod.eu/episodes/29/" rel="alternate" type="text/html" title="How Do We Know AI Gets Human Thinking Wrong? — Benjamin Paaßen" /><published>2025-09-19T00:00:00+00:00</published><updated>2025-09-19T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/29</id><content type="html" xml:base="https://deepdivepod.eu/episodes/29/"><![CDATA[]]></content><author><name></name></author><category term="Science" /><category term="Large Language Models" /><category term="Psychology" /><category term="Cognition" /><category term="Research Methods" /><category term="Evaluation and Testing" /><category term="Simulation" /><category term="Ethics" /><category term="Education" /><category term="Germany" /><summary type="html"><![CDATA[Host Iman Mossavat welcomes Dr. Benjamin Paaßen, Junior Professor for Knowledge Representation and Machine Learning at Bielefeld University and researcher at the Educational Technology Lab at DFKI. Together, they examine the growing debate around using large language models (LLMs) as substitutes for human participants in psychological and educational research. Dr. Paaßen shares why replacing human input with machine responses risks producing misleading results, especially when subtle linguistic shifts reveal critical differences between human and AI reasoning. He stresses that while LLMs can be useful in narrow, text-based tasks validated against large sets of human data, they cannot replicate the depth of human cognition. The discussion highlights the limits of AI in research settings, the dangers of overestimating machine intelligence, and the importance of maintaining rigorous standards when integrating these tools into scientific work.]]></summary></entry><entry><title type="html">Building Institutional Memory with AI: Enterprise Search — Jakub Zavrel</title><link href="https://deepdivepod.eu/episodes/28/" rel="alternate" type="text/html" title="Building Institutional Memory with AI: Enterprise Search — Jakub Zavrel" /><published>2025-09-03T00:00:00+00:00</published><updated>2025-09-03T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/28</id><content type="html" xml:base="https://deepdivepod.eu/episodes/28/"><![CDATA[]]></content><author><name></name></author><category term="Technology" /><category term="Large Language Models" /><category term="RAG" /><category term="Enterprise Search" /><category term="Knowledge Management" /><category term="Vector Databases" /><category term="AI Agents" /><category term="Regulation" /><category term="Future of Work" /><category term="Open Source" /><category term="Europe" /><category term="Startups" /><category term="Zeta Alpha" /><summary type="html"><![CDATA[Iman Mossavat speaks with Jakub Zavrel, CEO and founder of Zeta Alpha, about how AI can enhance enterprise search and knowledge management. Zavrel explains that large language models are limited unless connected to company-specific data, and that retrieval-augmented generation (RAG) systems, fine-tuned to organizational domains, improve accuracy and trust. He emphasizes the importance of data engineering, benchmarks, and secure access control, and reflects on the evolving landscape of open versus proprietary models. The conversation also touches on broader issues such as AI regulation, societal impacts, workforce augmentation, and the long-term vision of empowering organizations through AI-driven institutional memory.]]></summary></entry><entry><title type="html">Folded Narratives: AI, Protein Prediction and Scientific Silos — Daniel Cremers</title><link href="https://deepdivepod.eu/episodes/27/" rel="alternate" type="text/html" title="Folded Narratives: AI, Protein Prediction and Scientific Silos — Daniel Cremers" /><published>2025-08-10T00:00:00+00:00</published><updated>2025-08-10T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/27</id><content type="html" xml:base="https://deepdivepod.eu/episodes/27/"><![CDATA[]]></content><author><name></name></author><category term="Science" /><category term="Biology" /><category term="Protein Folding" /><category term="AlphaFold" /><category term="Deep Learning" /><category term="Machine Learning" /><category term="AI for Science" /><category term="Academia and Industry" /><category term="NeurIPS" /><category term="Europe" /><category term="Germany" /><summary type="html"><![CDATA[Iman Mossavat speaks with Professor Daniel Cremers, Director of the Munich Center for Machine Learning, Professor at the Technical University of Munich, and one of Germany's most distinguished AI researchers. Cremers is a recipient of the Gottfried Wilhelm Leibniz Prize and the 2024 ECCV Koenderink Test of Time Award. They discuss the groundbreaking work his team carried out in protein structure prediction back in 2016, years before AlphaFold made headlines. Although this work was presented at NeurIPS 2016, the flagship machine learning conference, in front of thousands of researchers, it was largely overlooked in subsequent citations. Beyond the science, Cremers reflects on the tension between academia and Big Tech, the challenges of recognition in a fast-moving field, and what it means to pursue meaningful work in AI today.]]></summary></entry><entry><title type="html">LSTM and the Rise of TiRex, the King of Time — Sepp Hochreiter</title><link href="https://deepdivepod.eu/episodes/26/" rel="alternate" type="text/html" title="LSTM and the Rise of TiRex, the King of Time — Sepp Hochreiter" /><published>2025-07-29T00:00:00+00:00</published><updated>2025-07-29T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/26</id><content type="html" xml:base="https://deepdivepod.eu/episodes/26/"><![CDATA[]]></content><author><name></name></author><category term="Technology" /><category term="Deep Learning" /><category term="LSTM" /><category term="xLSTM" /><category term="TiRex" /><category term="Time Series" /><category term="Recurrent Neural Networks" /><category term="Transformers" /><category term="Energy Efficiency" /><category term="Edge AI" /><category term="Open Source" /><category term="Europe" /><category term="Startups" /><category term="NXAI" /><category term="Austria" /><summary type="html"><![CDATA[The algorithm behind early versions of Siri, Google Translate, and Alexa? It wasn't invented by a tech giant. It was created by a master's student in the 1990s, rejected at first, then left to sleep quietly... until it changed the world. That student was Sepp Hochreiter, the inventor of LSTM, one of the most influential breakthroughs in AI. And now, he's back with TiRex, a time series foundation model that has outperformed others on Hugging Face time series leaderboard. It's fast, efficient, and built in Europe, where Sepp and his team at NXAI are betting on open source, smarter models, and AI that runs at the edge. In this episode, we explore how one of AI's pioneers is reshaping the future with fewer GPUs, smaller models, and a bold vision for European innovation.]]></summary></entry><entry><title type="html">Digital Detectives: The Science of Media Forensics — Tina Nikoukhah</title><link href="https://deepdivepod.eu/episodes/25/" rel="alternate" type="text/html" title="Digital Detectives: The Science of Media Forensics — Tina Nikoukhah" /><published>2025-07-20T00:00:00+00:00</published><updated>2025-07-20T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/25</id><content type="html" xml:base="https://deepdivepod.eu/episodes/25/"><![CDATA[]]></content><author><name></name></author><category term="Technology" /><category term="Mathematics" /><category term="Media Forensics" /><category term="Deepfakes" /><category term="Misinformation" /><category term="Watermarking" /><category term="C2PA" /><category term="Trust" /><category term="Media" /><category term="Startups" /><category term="GetReal" /><category term="France" /><summary type="html"><![CDATA[In this episode of Deep Dives with Iman, I speak with Dr. Tina Nikoukhah —applied mathematician, media forensics expert, and Director of Research at GetReal. We talk about how synthetic media is evolving, and how researchers like Tina are working to stay ahead of the curve. Tina shares why building public awareness is just as critical as tech solutions. (In fact, awareness and training are one of the four pillars of GetReal's approach, alongside real-time protection for live communication, media analysis tools, and support for media orgs.) It's a conversation about trust, technology, and how to navigate a digital world where seeing isn't always believing.]]></summary></entry><entry><title type="html">Perception as Inference — Hadi Vafaii</title><link href="https://deepdivepod.eu/episodes/24/" rel="alternate" type="text/html" title="Perception as Inference — Hadi Vafaii" /><published>2025-07-14T00:00:00+00:00</published><updated>2025-07-14T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/24</id><content type="html" xml:base="https://deepdivepod.eu/episodes/24/"><![CDATA[]]></content><author><name></name></author><category term="Science" /><category term="Mathematics" /><category term="Neuroscience" /><category term="Perception" /><category term="Cognition" /><category term="Bayesian Inference" /><category term="Uncertainty" /><category term="USA" /><summary type="html"><![CDATA[Can probability theory help explain how the mind works? This is the topic of this episode with Dr. Hadi Vafaii, a postdoctoral scholar at UC Berkeley’s Redwood Center for Theoretical Neuroscience, working in Jacob Yates’s lab. Hadi builds intuition step by step, using simple, relatable examples. He explains a mathematical perspective on how the brain processes information by inferring the unobserved causes behind what we see and hear, combining context with new evidence.]]></summary></entry><entry><title type="html">How Well-Meaning Leaders End Up at the Heart of Corporate Scandals — Guido Palazzo</title><link href="https://deepdivepod.eu/episodes/23/" rel="alternate" type="text/html" title="How Well-Meaning Leaders End Up at the Heart of Corporate Scandals — Guido Palazzo" /><published>2025-07-02T00:00:00+00:00</published><updated>2025-07-02T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/23</id><content type="html" xml:base="https://deepdivepod.eu/episodes/23/"><![CDATA[]]></content><author><name></name></author><category term="Society" /><category term="Ethics" /><category term="Business Ethics" /><category term="Ethical Blindness" /><category term="Corporate Scandals" /><category term="Business" /><category term="Psychology" /><category term="Regulation" /><category term="Switzerland" /><summary type="html"><![CDATA[In this episode, Dr. Guido Palazzo joins me to uncover the hidden dynamics behind corporate scandals. We discuss his latest book, The Dark Pattern, co-authored with Dr. Ulrich Hoffrage, which explores how even well-intentioned leaders can become ethically blind under intense organizational pressures. Through real-world cases from major global firms, Guido reveals how good people can end up doing bad things. He identifies nine recurring conditions that foster cultures where unethical behavior becomes normalized. Guido also shares his insights on how technology and AI can further deepen ethical lapses by adding layers of abstraction and distance.]]></summary></entry><entry><title type="html">The True Value of AI Lies in Testing and Evaluation — Shea Brown</title><link href="https://deepdivepod.eu/episodes/22/" rel="alternate" type="text/html" title="The True Value of AI Lies in Testing and Evaluation — Shea Brown" /><published>2025-06-23T00:00:00+00:00</published><updated>2025-06-23T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/22</id><content type="html" xml:base="https://deepdivepod.eu/episodes/22/"><![CDATA[]]></content><author><name></name></author><category term="Society" /><category term="Accountability" /><category term="Auditing" /><category term="Evaluation and Testing" /><category term="Governance" /><category term="Regulation" /><category term="Bias and Fairness" /><category term="Trust" /><category term="Risk" /><category term="Future of Work" /><category term="Startups" /><category term="Physics" /><category term="USA" /><summary type="html"><![CDATA[I got the opportunity to talk to Dr. Shea Brown, former astrophysics professor, and founder of BABL AI, for a sharp look at AI accountability. Shea shares his journey from studying the cosmos to interrogating AI systems, pointing out how scientific skepticism must carry over into today's tech landscape. At BABL AI, he leads efforts to audit and govern AI to ensure it serves human interests, especially in high-stakes areas like hiring, credit scoring, and public services. With real-world examples and hands-on audit methods, Shea explains how AI systems often fail silently and how risk-based auditing can surface hidden harms. Shea hopes for a future where regulation, education, and responsible innovation work hand in hand for human flourishing.]]></summary></entry><entry><title type="html">AI for School Selection: Efficient, Fair or Transparent? Pick Two — Mayesha Tasnim &amp;amp; Paul Verhagen</title><link href="https://deepdivepod.eu/episodes/21/" rel="alternate" type="text/html" title="AI for School Selection: Efficient, Fair or Transparent? Pick Two — Mayesha Tasnim &amp;amp; Paul Verhagen" /><published>2025-06-20T00:00:00+00:00</published><updated>2025-06-20T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/21</id><content type="html" xml:base="https://deepdivepod.eu/episodes/21/"><![CDATA[]]></content><author><name></name></author><category term="Society" /><category term="Education" /><category term="Bias and Fairness" /><category term="Transparency" /><category term="Public Policy" /><category term="Game Theory" /><category term="Civic AI Lab" /><category term="Amsterdam" /><category term="Eindhoven" /><category term="Netherlands" /><summary type="html"><![CDATA[I explore the complex intersection of algorithms and education policy with Mayesha Tasnim and Paul Verhagen of the Civic AI Lab at the University of Amsterdam. In cities like Amsterdam and Eindhoven, school placement is increasingly driven by algorithms. But how fair and transparent are these systems? Together, we unpack how public algorithms shape student futures, revealing the trade-offs between fairness, efficiency, and transparency. From the role of lotteries in Amsterdam's admissions to the challenges of "gaming the system," this episode dives into what it truly means to design equitable technology in public policy. Listeners gain rare insight into how smarter algorithms can improve outcomes, while still risking unintended consequences.]]></summary></entry><entry><title type="html">Where Geometry Meets Mission — Colleen Farrelly</title><link href="https://deepdivepod.eu/episodes/20/" rel="alternate" type="text/html" title="Where Geometry Meets Mission — Colleen Farrelly" /><published>2025-06-08T00:00:00+00:00</published><updated>2025-06-08T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/20</id><content type="html" xml:base="https://deepdivepod.eu/episodes/20/"><![CDATA[]]></content><author><name></name></author><category term="Science" /><category term="Mathematics" /><category term="Geometry" /><category term="Topology" /><category term="Modelling" /><category term="Biology" /><category term="Healthcare" /><category term="Epidemiology" /><category term="Generative AI" /><category term="USA" /><summary type="html"><![CDATA[Colleen Farrelly uses mathematics as a versatile tool across many fields, driven by curiosity and social purpose. Calling herself a “wandering mathematician,” she connects pure math with real-world problems—from epidemic modeling and radiomics to data analysis. Colleen shares her experience modeling to help position resources that limited an Ebola outbreak. She explains how understanding the geometry and topology of medical images helps generative algorithms produce medically plausible data. She also highlights how tools useful in one domain, like finance, can be applied to others such as ecology or marketing. She stresses the importance of critical thinking: questioning and exploring better approaches.]]></summary></entry><entry><title type="html">No One Left Behind: Making Math a Human Experience — Luis Serrano</title><link href="https://deepdivepod.eu/episodes/19/" rel="alternate" type="text/html" title="No One Left Behind: Making Math a Human Experience — Luis Serrano" /><published>2025-06-01T00:00:00+00:00</published><updated>2025-06-01T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/19</id><content type="html" xml:base="https://deepdivepod.eu/episodes/19/"><![CDATA[]]></content><author><name></name></author><category term="Science" /><category term="Mathematics" /><category term="Education" /><category term="Machine Learning" /><category term="Quantum Computing" /><category term="Emotions and Empathy" /><category term="Serrano Academy" /><category term="Canada" /><summary type="html"><![CDATA[Luis Serrano, founder of Serrano Academy, YouTuber, and author of Grokking Machine Learning, joins Deep Dives with Iman to unpack how we teach, learn, and think about mathematics. Luis brings a heartfelt, vulnerable, and deeply thoughtful philosophy to STEM education. Luis's philosophy is shaped by his journey from mathematician to machine learning engineer at Google, to AI educator at Udacity and Apple, to quantum researcher at Zapata, and now founder of Serrano Academy.]]></summary></entry><entry><title type="html">Embodied Cognition, Dynamical Systems &amp;amp; Attractor Networks in Care Robotics — Yulia Sandamirskaya</title><link href="https://deepdivepod.eu/episodes/18/" rel="alternate" type="text/html" title="Embodied Cognition, Dynamical Systems &amp;amp; Attractor Networks in Care Robotics — Yulia Sandamirskaya" /><published>2025-05-19T00:00:00+00:00</published><updated>2025-05-19T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/18</id><content type="html" xml:base="https://deepdivepod.eu/episodes/18/"><![CDATA[]]></content><author><name></name></author><category term="Technology" /><category term="Science" /><category term="Robotics" /><category term="Neuromorphic Computing" /><category term="Neuroscience" /><category term="Energy Efficiency" /><category term="Embodied Cognition" /><category term="Dynamical Systems" /><category term="Attractor Networks" /><category term="Switzerland" /><summary type="html"><![CDATA[In this episode of Deep Dives with Iman, I speak with Prof. Dr. Yulia Sandamirskaya, head of the Center for Cognitive Computing in Life Sciences at Zurich University of Applied Sciences. Yulia builds robots that perceive, plan, and act in the real world using data- and energy-efficient neuromorphic computing. She states that we discovered our brain didn’t evolve for abstract tasks, it evolved to move. Even small children show cognitive behaviors grounded in movement. [...] Like when we were little, doing simple addition- ‘seven plus two’ - we imagine seven on a number line and move right two units. We place things in space.]]></summary></entry><entry><title type="html">Safe AI for Children — Tara Steele</title><link href="https://deepdivepod.eu/episodes/17/" rel="alternate" type="text/html" title="Safe AI for Children — Tara Steele" /><published>2025-05-09T00:00:00+00:00</published><updated>2025-05-09T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/17</id><content type="html" xml:base="https://deepdivepod.eu/episodes/17/"><![CDATA[]]></content><author><name></name></author><category term="Society" /><category term="Ethics" /><category term="Children" /><category term="AI Safety" /><category term="Grief Bots" /><category term="Chatbots" /><category term="Safe AI for Children Alliance" /><category term="UK" /><summary type="html"><![CDATA[I had the privilege of speaking with Tara Steele, Founder and Director of the Safe AI for Children Alliance. We discussed the growing concerns around AI’s impact on children, particularly with the rise of grief bots—chatbots designed to simulate deceased people—and the lack of understanding regarding their effects.]]></summary></entry><entry><title type="html">Seeing Atoms, Shaping Futures: AI, Science, Purpose and Strategy — Remco Schoenmakers</title><link href="https://deepdivepod.eu/episodes/16/" rel="alternate" type="text/html" title="Seeing Atoms, Shaping Futures: AI, Science, Purpose and Strategy — Remco Schoenmakers" /><published>2025-04-10T00:00:00+00:00</published><updated>2025-04-10T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/16</id><content type="html" xml:base="https://deepdivepod.eu/episodes/16/"><![CDATA[]]></content><author><name></name></author><category term="Science" /><category term="Physics" /><category term="Microscopy" /><category term="Materials Science" /><category term="AI for Science" /><category term="Academia and Industry" /><category term="Future of Work" /><category term="Human-AI Collaboration" /><category term="Thermo Fisher Scientific" /><category term="Eindhoven" /><category term="Brainport" /><summary type="html"><![CDATA[In this episode of Deep Dives with Iman, I am joined by Dr. Remco Schoenmakers, Senior Director and AI Strategy Lead at Thermo Fisher Scientific. Dr. Schoenmakers discusses his journey in AI, from his early days in astrophysics to leading AI strategy for the Electron Microscopy business. He shares insights on how AI is transforming the scientific landscape, emphasizing that it is a tool to enhance, not replace, human expertise. With his vast experience, Dr. Schoenmakers also addresses the growing influence of AI in workplaces, offering perspectives on the role of future generations of scientists.]]></summary></entry><entry><title type="html">AI, Art and the Human Touch: Creativity in the Age of Machines — Mei-Li Nieuwland</title><link href="https://deepdivepod.eu/episodes/15/" rel="alternate" type="text/html" title="AI, Art and the Human Touch: Creativity in the Age of Machines — Mei-Li Nieuwland" /><published>2025-03-17T00:00:00+00:00</published><updated>2025-03-17T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/15</id><content type="html" xml:base="https://deepdivepod.eu/episodes/15/"><![CDATA[]]></content><author><name></name></author><category term="Society" /><category term="Creativity" /><category term="Art" /><category term="Illustration" /><category term="Design" /><category term="Culture" /><category term="Generative AI" /><category term="Future of Work" /><summary type="html"><![CDATA[How is AI shaping the future of creativity? Where does human artistry remain irreplaceable? Mei-Li is an incredible illustrator and graphic journalist with a background in AI and cultural anthropology. In our conversation, we explored AI's growing influence on the art industry, its impact on concept and environmental art, and why her niche remains largely unaffected.]]></summary></entry><entry><title type="html">Could Generative AI Be Useful for Science and Understanding the World? — Jakub Tomczak</title><link href="https://deepdivepod.eu/episodes/14/" rel="alternate" type="text/html" title="Could Generative AI Be Useful for Science and Understanding the World? — Jakub Tomczak" /><published>2025-03-05T00:00:00+00:00</published><updated>2025-03-05T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/14</id><content type="html" xml:base="https://deepdivepod.eu/episodes/14/"><![CDATA[]]></content><author><name></name></author><category term="Science" /><category term="Mathematics" /><category term="Generative AI" /><category term="Machine Learning" /><category term="AI for Science" /><category term="Biology" /><category term="Drug Discovery" /><category term="Large Language Models" /><category term="USA" /><summary type="html"><![CDATA[In this episode, Jakub Tomczak, a leading figure in Generative AI and former Program Chair of NeurIPS 2024, shares his insights on the transformative impact of Generative AI, not just in the field of artificial intelligence but in scientific discovery as well. Jakub explains how this technology is reshaping our understanding of generation processes and its potential to revolutionize various domains. Jakub offers a thought-provoking perspective on AI's societal implications, inviting us to be mindful. For those interested in a deeper dive into his work, Jakub's book, Deep Generative Modeling, is an essential resource on advanced AI models, including diffusion, flow, and energy-based models, GANs, and Variational Autoencoders.]]></summary></entry><entry><title type="html">Your Business Opportunity Is Trustworthiness — Ger Janssen</title><link href="https://deepdivepod.eu/episodes/13/" rel="alternate" type="text/html" title="Your Business Opportunity Is Trustworthiness — Ger Janssen" /><published>2025-02-02T00:00:00+00:00</published><updated>2025-02-02T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/13</id><content type="html" xml:base="https://deepdivepod.eu/episodes/13/"><![CDATA[]]></content><author><name></name></author><category term="Society" /><category term="Ethics" /><category term="Responsible AI" /><category term="Regulation" /><category term="Bias and Fairness" /><category term="Trust" /><category term="Transparency" /><category term="Human-AI Collaboration" /><category term="AI Literacy" /><category term="Healthcare" /><category term="Synthetic Data" /><category term="Large Language Models" /><category term="Generative AI" /><category term="Philips" /><category term="Industry" /><category term="Eindhoven" /><summary type="html"><![CDATA[Together with Ger Janssen, we discuss responsible AI practices, highlighting trustworthiness, fairness, and transparency in AI applications. Ger, Philips' Ethics and Compliance Lead, explores AI’s impact on industries, particularly healthcare. He discusses how AI can enhance patient care while addressing biases and ethical challenges. With AI's rapid rise, it’s crucial to adapt education and regulations to support effective human-AI collaboration. As Janssen underscores, AI isn’t going away—businesses must learn to leverage it responsibly. This episode offers essential insights into AI’s evolving influence on industries and society.]]></summary></entry><entry><title type="html">How Can Nature Inspire Energy-Efficient AI? — Federico Corradi</title><link href="https://deepdivepod.eu/episodes/12/" rel="alternate" type="text/html" title="How Can Nature Inspire Energy-Efficient AI? — Federico Corradi" /><published>2025-01-26T00:00:00+00:00</published><updated>2025-01-26T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/12</id><content type="html" xml:base="https://deepdivepod.eu/episodes/12/"><![CDATA[]]></content><author><name></name></author><category term="Technology" /><category term="Science" /><category term="Neuromorphic Computing" /><category term="Neuroscience" /><category term="Energy Efficiency" /><category term="Edge AI" /><category term="Hardware" /><category term="In-Memory Computing" /><category term="Large Language Models" /><category term="Robotics" /><category term="Eindhoven" /><summary type="html"><![CDATA[A brief introduction to neuromorphic computing by Dr. Federico Corradi, Assistant Professor at TU Eindhoven. Specializing in energy-efficient AI inspired by the brain, Dr. Corradi leads the Neuromorphic Edge Computing Systems Lab, exploring how nature's principles can reshape AI to consume less energy. In this episode, Dr. Corradi explains how the brain’s minimal energy use inspires new AI systems, addressing the massive energy demands of large language models. In his line of research, the boundary between hardware and software is increasingly blurred. Designing algorithms and hardware together could transform AI into more sustainable and independent systems, enabling smarter edge devices without cloud reliance.]]></summary></entry><entry><title type="html">Agents Need a Sense of Intent and Purpose — Robert Engels</title><link href="https://deepdivepod.eu/episodes/11/" rel="alternate" type="text/html" title="Agents Need a Sense of Intent and Purpose — Robert Engels" /><published>2025-01-13T00:00:00+00:00</published><updated>2025-01-13T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/11</id><content type="html" xml:base="https://deepdivepod.eu/episodes/11/"><![CDATA[]]></content><author><name></name></author><category term="Technology" /><category term="Large Language Models" /><category term="AI Agents" /><category term="Reasoning" /><category term="Abstraction" /><category term="Context" /><category term="Philosophy" /><category term="Self-Driving Cars" /><category term="Capgemini" /><category term="Eindhoven" /><summary type="html"><![CDATA[Iman Mossavat engages Robert Engels, Head of AI Futures Lab at Capgemini, in a discussion on the role of context and abstraction in AI. With 36 years of experience, Engels examines why current generative AI systems, like GPT-4, excel at tasks yet fail in complex, real-world settings. Robert argues, “Two things are underperforming in the world of generative AI—abstraction and logical reasoning.” Robert underscores the need for AI to adopt a world model akin to philosophical reasoning: "Plato and Socrates understood this when they built logics—they looked at the world and tried to grasp its principles."]]></summary></entry><entry><title type="html">AI vs Doctors: Human Judgement Wins (For Now) in Complex Medical Contexts — Artin Entezarjou</title><link href="https://deepdivepod.eu/episodes/10/" rel="alternate" type="text/html" title="AI vs Doctors: Human Judgement Wins (For Now) in Complex Medical Contexts — Artin Entezarjou" /><published>2025-01-04T00:00:00+00:00</published><updated>2025-01-04T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/10</id><content type="html" xml:base="https://deepdivepod.eu/episodes/10/"><![CDATA[]]></content><author><name></name></author><category term="Science" /><category term="Healthcare" /><category term="Medicine" /><category term="Diagnosis" /><category term="Large Language Models" /><category term="Reasoning" /><category term="Evaluation and Testing" /><category term="Human-AI Collaboration" /><category term="Decision Making" /><category term="Sweden" /><summary type="html"><![CDATA[Our host Iman Mossavat welcomes Dr. Artin Entezarjou, a board-certified general medicine specialist, to discuss AI's role in healthcare. Dr. Entezarjou highlights AI's progress and challenges in handling complex clinical scenarios. “Humans can outperform AI when questions aren’t multiple choice, especially when psychosocial factors are involved,” he explains. His research comparing GPT-4 to human doctors reveals AI's limitations in understanding real-world medical complexities and patient contexts. “Clinical decisions require more than symptoms; they demand insight into patients’ preferences and circumstances,” he says. Dr. Entezarjou stresses the continued need for human judgment, adding, “We’re still the masters of recognizing when more context is needed, though this is rapidly changing.” He advocates for AI systems that are robust, trustworthy, and intuitive, emphasizing their role in supporting—not replacing—physicians. “AI can excel in specific tasks, but general judgment remains human,” he concludes.]]></summary></entry><entry><title type="html">What If We Optimized for Enough, Not More? — Diederik Roijers</title><link href="https://deepdivepod.eu/episodes/9/" rel="alternate" type="text/html" title="What If We Optimized for Enough, Not More? — Diederik Roijers" /><published>2024-12-16T00:00:00+00:00</published><updated>2024-12-16T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/9</id><content type="html" xml:base="https://deepdivepod.eu/episodes/9/"><![CDATA[]]></content><author><name></name></author><category term="Technology" /><category term="Reinforcement Learning" /><category term="Optimization" /><category term="Multi-Objective Optimization" /><category term="Human Values" /><category term="Ethics" /><category term="Accountability" /><category term="AI Agents" /><category term="Decision Making" /><summary type="html"><![CDATA[Join Iman Mossavat as he speaks with Diederik Roijers, senior researcher at the AI Lab at Vrije Universiteit Brussel. Diederik specializes in Multi-Objective Reinforcement Learning (MORL), a framework that shifts focus from optimizing a single outcome to balancing multiple, sometimes competing, objectives. Drawing on the ancient idea of balance—akin to Yin and Yang—Diederik challenges the status quo of “maximizing more.” Instead, he advocates for pursuing outcomes that are "good enough", prioritizing practicality, ethics, and societal benefit over perfection. 💡 Key Insight: How can AI navigate trade-offs like maximizing rewards while minimizing risks? Diederik’s vision emphasizes deliberate choices in AI design, promoting transparency, maintainability, and alignment with societal values. Together, they explore how AI can serve everyone fairly—not just a select few—while balancing innovation and responsibility. Tune in for a refreshing perspective on building AI systems that prioritize balance, fairness, and shared benefit.]]></summary></entry><entry><title type="html">Lazy But Brilliant: Real-Time Decision Making with Reactive AI — Albert Podusenko &amp;amp; İsmail Şenöz</title><link href="https://deepdivepod.eu/episodes/7/" rel="alternate" type="text/html" title="Lazy But Brilliant: Real-Time Decision Making with Reactive AI — Albert Podusenko &amp;amp; İsmail Şenöz" /><published>2024-12-04T00:00:00+00:00</published><updated>2024-12-04T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/7</id><content type="html" xml:base="https://deepdivepod.eu/episodes/7/"><![CDATA[]]></content><author><name></name></author><category term="Technology" /><category term="Software Engineering" /><category term="Reliability" /><category term="RxInfer" /><category term="Probabilistic Programming" /><category term="Bayesian Inference" /><category term="Uncertainty" /><category term="AI Agents" /><category term="Decision Making" /><category term="Open Source" /><category term="Startups" /><category term="Legal Tech" /><summary type="html"><![CDATA[Join Iman Mossavat for an insightful episode with special guests Albert Podusenko and İsmail Şenöz, as we dive into the innovative world of Lazy Dynamics. This cutting-edge company is reshaping how agents process and act on real-time data in unpredictable environments. Their software streamlines the development of intelligent systems capable of navigating uncertainty — whether it's an ambulance optimizing routes or legal tech helping lawyers model strategic uncertainties in complex litigation. At the heart of their approach is RxInfer, a fast and efficient tool designed to overcome the computational limitations of traditional probabilistic programming libraries, enabling real-time decision-making. With a unique reactive message-passing system, Lazy Dynamics ensures agents can keep reasoning even if some sensors fail — a true breakthrough for dynamic, high-stakes environments. Discover how collaboration and open-source contributions are fueling their success and how these innovations are shaping the future. Tune in for an episode packed with insights into the future of AI and real-time decision-making! For more about the guests: Albert Podusenko on Google Scholar İsmail Şenöz on Google Scholar]]></summary></entry><entry><title type="html">Weaving Trustworthy AI with Threads of Reason — Mehdi Dastani</title><link href="https://deepdivepod.eu/episodes/8/" rel="alternate" type="text/html" title="Weaving Trustworthy AI with Threads of Reason — Mehdi Dastani" /><published>2024-12-03T00:00:00+00:00</published><updated>2024-12-03T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/8</id><content type="html" xml:base="https://deepdivepod.eu/episodes/8/"><![CDATA[]]></content><author><name></name></author><category term="Technology" /><category term="AI Agents" /><category term="Reasoning" /><category term="Philosophy" /><category term="Psychology" /><category term="Human Values" /><category term="AI Safety" /><category term="Accountability" /><category term="Decision Making" /><category term="Generative AI" /><summary type="html"><![CDATA[Join Iman Mossavat for an enlightening conversation with Professor Mehdi Dastani, Chair of the Intelligent Systems group at Utrecht University. With over 500 scientific contributions, Professor Dastani has spent decades merging computer science and philosophy in a way that challenges and inspires. His work spans formal logic, reinforcement learning, ethics, and human-centered AI, bringing fresh perspectives to the challenges facing modern technology. Key Insight: Professor Dastani emphasizes the need for a fusion of machine learning, formal reasoning, and domain expertise to create AI systems that are safe, ethical, and aligned with human values. His interdisciplinary approach, drawing from psychology, law, and philosophy, offers innovative solutions to issues like AI bias, accountability, and safety risks. Mehdi Dastani on Google Scholar]]></summary></entry><entry><title type="html">Is Your AI Project Legally Compliant? — Inge Brattinga</title><link href="https://deepdivepod.eu/episodes/6/" rel="alternate" type="text/html" title="Is Your AI Project Legally Compliant? — Inge Brattinga" /><published>2024-11-15T00:00:00+00:00</published><updated>2024-11-15T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/6</id><content type="html" xml:base="https://deepdivepod.eu/episodes/6/"><![CDATA[]]></content><author><name></name></author><category term="Society" /><category term="Law" /><category term="Ethics" /><category term="Regulation" /><category term="Bias and Fairness" /><category term="Accountability" /><category term="Future of Work" /><category term="AI Literacy" /><category term="Decision Making" /><summary type="html"><![CDATA[In this episode, Inge Brattinga, lawyer at VRF Advocaten and lecturer at Avans University, discusses how AI is reshaping hiring, human-resources (HR), decision-making, and privacy. She highlights the importance of AI literacy, offering practical advice for businesses navigating upcoming regulations.]]></summary></entry><entry><title type="html">Evolutionary Science and AI — Indre Žliobaitė</title><link href="https://deepdivepod.eu/episodes/5/" rel="alternate" type="text/html" title="Evolutionary Science and AI — Indre Žliobaitė" /><published>2024-10-18T00:00:00+00:00</published><updated>2024-10-18T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/5</id><content type="html" xml:base="https://deepdivepod.eu/episodes/5/"><![CDATA[]]></content><author><name></name></author><category term="Science" /><category term="Evolution" /><category term="Biology" /><category term="Machine Learning" /><category term="Concept Drift" /><category term="Data Science" /><summary type="html"><![CDATA[Professor Indre Žliobaitė explores parallels between evolutionary science and AI, discussing adaptive models, concept drift, and more. She explains how principles of evolution can inform AI, providing insights into dynamics like competition and adaptation.]]></summary></entry><entry><title type="html">Legal Challenges of AI — Colette Cuijpers</title><link href="https://deepdivepod.eu/episodes/4/" rel="alternate" type="text/html" title="Legal Challenges of AI — Colette Cuijpers" /><published>2024-09-09T00:00:00+00:00</published><updated>2024-09-09T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/4</id><content type="html" xml:base="https://deepdivepod.eu/episodes/4/"><![CDATA[]]></content><author><name></name></author><category term="Society" /><category term="Law" /><category term="Regulation" /><category term="Privacy" /><category term="Accountability" /><category term="Chatbots" /><category term="Large Language Models" /><summary type="html"><![CDATA[Colette Cuijpers, Associate Professor at Tilburg Law School, discusses the urgent legal and ethical challenges AI presents, from accountability to bias. This episode highlights the balance between innovation and regulation in a rapidly evolving field.]]></summary></entry><entry><title type="html">Causal AI and Intelligent Systems — Alexander Molak</title><link href="https://deepdivepod.eu/episodes/3/" rel="alternate" type="text/html" title="Causal AI and Intelligent Systems — Alexander Molak" /><published>2024-08-22T00:00:00+00:00</published><updated>2024-08-22T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/3</id><content type="html" xml:base="https://deepdivepod.eu/episodes/3/"><![CDATA[]]></content><author><name></name></author><category term="Technology" /><category term="Causal Inference" /><category term="Reasoning" /><category term="Bayesian Inference" /><category term="Philosophy" /><summary type="html"><![CDATA[Alexander Molak, machine learning researcher, explains why Generative AI lacks true causal understanding. He discusses Causal AI and its potential to improve intelligent systems by moving beyond mere correlation to deeper cause-and-effect insights.]]></summary></entry><entry><title type="html">Systems Engineering and AI — Gerrit Muller</title><link href="https://deepdivepod.eu/episodes/2/" rel="alternate" type="text/html" title="Systems Engineering and AI — Gerrit Muller" /><published>2024-07-25T00:00:00+00:00</published><updated>2024-07-25T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/2</id><content type="html" xml:base="https://deepdivepod.eu/episodes/2/"><![CDATA[]]></content><author><name></name></author><category term="Technology" /><category term="Systems Engineering" /><category term="Complexity" /><category term="Design" /><summary type="html"><![CDATA[Gerrit Muller, a systems engineer and professor, discusses the integration of systems engineering and AI, exploring the importance of a balanced approach. He examines common challenges, such as data quality and interpretability, and the complementary role of Model-Based Systems Engineering.]]></summary></entry><entry><title type="html">Understanding Power Dynamics in AI — Mahault Albarracin</title><link href="https://deepdivepod.eu/episodes/1/" rel="alternate" type="text/html" title="Understanding Power Dynamics in AI — Mahault Albarracin" /><published>2024-04-30T00:00:00+00:00</published><updated>2024-04-30T00:00:00+00:00</updated><id>https://deepdivepod.eu/episodes/1</id><content type="html" xml:base="https://deepdivepod.eu/episodes/1/"><![CDATA[]]></content><author><name></name></author><category term="Society" /><category term="Ethics" /><category term="Power" /><summary type="html"><![CDATA[Mahault Albarracin, Director of Applied Research and PhD candidate in Computing, discusses how AI often prioritizes objectives set by powerful stakeholders, affecting societal power structures. This episode explores the intersection of technology, ethics, and power in AI design.]]></summary></entry></feed>