Christian Guttmann

Interesting paper alert - a commentary in Nature Medicine highlights a pivotal tension in healthcare today: while #AIagents promise transformative impacts — from autonomous triage to multi-modal clinical decision support — regulatory frameworks remain built for a different era (a recurring scheme in the world of Health AI which I do not see going away anytime soon: technical progress in AI will always outpace regulatory actions). Here are the take aways from the paper: 🔹 Current state of regulation of AI agents in health and medicine: Most regulatory pathways, like the EU Medical Device Regulation and the U.S. FDA’s 510(k), are optimized for narrow, static, non-adaptive tools. Approval of AI-driven devices with broad scope and high autonomy remains rare and challenging. 🔹 The problem (why existing regulatory frameworks fall short of capturing agentic AI applications): AI agents orchestrate multiple tools, access sensitive data, and make goal-directed decisions autonomously. Their adaptive nature, coupled with non-deterministic outputs and complex multi-component architectures, pushes them outside the assumptions underpinning today’s medical device regulation. 🔹 Proposed solutions (how can we explore and experiment safely with AI agents in clinical settings in the absence of comprehensive regulatory guidance): The paper explores several emerging regulatory strategies, including: Voluntary alternative pathways → tailored approvals for high-risk AI agents Adaptive oversight → dynamic, real-world performance monitoring Regulatory sandboxes → controlled environments to experiment safely Outcome-driven evaluation → focusing on patient-centered results rather than static technical compliance 💡 The takeaway (so where does this leave us): While regulation is still playing catch-up with the risks and complexities of AI agents in healthcare, it is possible to safely experiment and sandbox their use — provided there are clear guardrails, strong oversight, and rigorous evaluation frameworks. This balanced, sandbox-first approach could enable innovation without compromising patient safety. For first hand insights on how this is all playing out in the real world right now, have a read of these two recent papers from Eric Topol, MD (https://lnkd.in/gEycp__S) and Harsha Nori (https://lnkd.in/gKNxpxsA) John Lim Christopher Bain ZongYuan 宗元 Ge 戈 Deval Mehta #AI #Healthcare #Regulation #AIagents #DigitalHealth #NatureMedicine

136

13 kommentarer

🚀 Big week ahead for the Quantum + AI team at IEEE Quantum Week 2025! We’re thrilled to share our latest work at the intersection of Quantum Computing and Artificial Intelligence. If you're attending, here’s where you can find us: - 📅 Sunday, August 31 – 3:00PM 🧠 Paper Presentation Juan Cruz-Benito presents, "Quantum Processing Unit (QPU) Processing Time Prediction with Machine Learning" - How ML can improve QPU time prediction for quantum jobs—enabling smarter quantum workload scheduling. - 📅 Monday, September 1 – 10:00AM 🛠️ Tutorial: TUT13 — AI Methods for Quantum Circuit Optimization David Kremer and Juan Cruz-Benito will demonstrate how Reinforcement Learning can be used to transpile and synthesize quantum circuits, using qiskit-ibm-transpiler and other tools developed by our team. - 📅 Tuesday, September 2 – 12:00PM💡 Live Demo at IBM Quantum Booth #400 Juan Cruz-Benito will showcase our new AI-powered quantum computing development environment—designed to visually bridge the gap between quantum theory and practical implementation. - 📅 Thursday, September 4 – 10:00AM 🎤 Invited Talk at the 3rd Workshop on Quantum Algorithm Design Automation Nicolas Dupuis  presents: "Large Language Models for Generating Quantum Computing Code" - Learn how we’re using QPUs for post-training LLMs to power the next generation of Quantum and Qiskit Code Assistants. We’re looking forward to connecting with the community, exchanging ideas, and pushing the boundaries of what’s possible with Quantum + AI. If you're attending, come say hi! #QuantumComputing #AI #MachineLearning #Qiskit #IEEEQuantumWeek #QuantumAI #ReinforcementLearning #IBMQuantum

110

1 kommentar

How do #AI models learn to behave not just predict? Large language models like #ChatGPT, #Claude, and #LLaMA are getting smarter every day. But it’s not enough for them to just generate text, they need to align with human values, tone, and intent. That’s where Reinforcement Learning from Human Feedback (#RLHF) comes in. It’s the secret sauce behind today's most aligned, helpful, and agentic AI systems. If you're building with LLMs or exploring AI alignment, this guide is for you. Read the full blog here: https://lnkd.in/gW4JMgMA #RLHF #AIAlignment #AgenticAI #LLM #MachineLearning #GenerativeAI #NLP #OpenAI #Anthropic #ResponsibleAI #DeepLearning

36

6 kommentarer

LLMs can do it all, so why build internal AI teams? AI hiring freezes are back in the headlines and some suggest that reason is that we don’t need more AI hires, LLM’s will do it all. Here’s a quick 2-lens reality check: 1- Task Fit Not all tasks are general-purpose. Diagnosing brain tumors ≠ summarizing emails. A recent MIT-BCG study (2025) found that 95% of failed GenAI pilots lacked domain adaptation. LLMs trained on internet-scale data struggle in high-stakes, specialized domains. 2- Model Limits Yann LeCun and many leading AI voices have said transformers are a stepping stone, not a summit. They lack reasoning, grounding, and efficiency. exactly what specialized use cases require. Meta’s 2025 JEPA and DeepMind’s Gemini are already exploring post-transformer paths. Bottom line: LLMs are powerful, but not a silver bullet. AI isn’t just a plug-and-play engineering tool, it’s a field still under construction. #AI #LLMs #DeepTech #SpecializedAI #HealthTech #MIT #MetaAI #Gemini #JEPA

32

🚀 𝗕𝘂𝗶𝗹𝗱𝗶𝗻𝗴 𝗔𝗜 𝗣𝗿𝗼𝗱𝘂𝗰𝘁𝘀 𝗶𝗻 𝘁𝗵𝗲 𝗣𝗿𝗼𝗯𝗮𝗯𝗶𝗹𝗶𝘀𝘁𝗶𝗰 𝗘𝗿𝗮 (a great blog post) We’ve left deterministic software behind. Traditional systems were predictable: the same input gave the same output. Easy to measure, easy to plan. AI doesn’t work that way. It runs on 𝗽𝗿𝗼𝗯𝗮𝗯𝗶𝗹𝗶𝘁𝘆, 𝘂𝗻𝗰𝗲𝗿𝘁𝗮𝗶𝗻𝘁𝘆, 𝗮𝗻𝗱 𝗲𝗺𝗲𝗿𝗴𝗲𝗻𝗰𝗲. Outputs aren’t fixed — they’re distributions. Funnels, rigid dashboards, and old SLOs can’t capture that complexity anymore. This shift forces us to rethink how we build: •  𝗣𝗹𝗮𝘆𝗯𝗼𝗼𝗸𝘀 𝗮𝗿𝗲 𝗼𝗯𝘀𝗼𝗹𝗲𝘁𝗲. Product, marketing, and engineering can’t operate in silos when the system itself is probabilistic. •  𝗗𝗮𝘁𝗮 𝗶𝘀 𝘁𝗵𝗲 𝗻𝗲𝘄 𝗰𝗼𝗺𝗽𝗮𝘀𝘀. Instead of forcing reality into neat funnels, we need to observe, measure, and adapt to what actually happens. • 𝗘𝗺𝗲𝗿𝗴𝗲𝗻𝗰𝗲 𝗶𝘀 𝘁𝗵𝗲 𝗼𝗽𝗽𝗼𝗿𝘁𝘂𝗻𝗶𝘁𝘆. The magic of AI isn’t in deterministic predictions, it’s in new behaviors that show up when models and users interact. The future belongs to teams that 𝘁𝗵𝗶𝗻𝗸 𝗽𝗿𝗼𝗯𝗮𝗯𝗶𝗹𝗶𝘀𝘁𝗶𝗰𝗮𝗹𝗹𝘆, 𝗱𝗲𝘀𝗶𝗴𝗻 𝗳𝗼𝗿 𝘂𝗻𝗰𝗲𝗿𝘁𝗮𝗶𝗻𝘁𝘆, and 𝘁𝗿𝗲𝗮𝘁 𝗱𝗮𝘁𝗮 𝗮𝘀 𝘁𝗵𝗲 𝗺𝗮𝗽 — 𝗻𝗼𝘁 𝘁𝗵𝗲 𝗮𝗳𝘁𝗲𝗿𝘁𝗵𝗼𝘂𝗴𝗵𝘁. 👉 Great piece by Gian Segato: https://lnkd.in/ggC-DYJ5

36

1 kommentar

The adoption of LLMs in healthcare is quick, but several important questions remain. One we addressed in our latest work is: are we paying enough attention to the subtle choices that determine their reliability?  While the focus is often on the model itself, a seemingly minor technical detail, how the model chooses its next word, can alter the quality of its output. In "A Comparative Study of Decoding Strategies in Medical Text Generation" led by Oriana Presacan, we systematically investigated how 11 different decoding strategies impact the performance of both general-purpose and medically-specialized LLMs across five distinct healthcare tasks. Full paper: https://lnkd.in/df9AjssD Code and dataset: https://lnkd.in/dF5kuuHb Our findings challenge some common assumptions about deploying LLMs in high-stakes environments: -The decoding strategy is not a minor detail; it is an important choice. We found that deterministic strategies like beam search consistently produce higher quality and more reliable outputs than common stochastic methods like top-k sampling. The choice of strategy can be as significant as the choice of the LLM itself. -Specialized medical LLMs are surprisingly sensitive. Contrary to the belief and discussions that domain-specific models are inherently more robust, our results show they are significantly more sensitive to the selected decoding method than their general-purpose counterparts. A medical model that performs well with one strategy can see its performance degrade significantly with another. -There is also a direct trade-off between speed and quality. Our analysis revealed a statistically significant positive correlation between inference time and output quality. In medical applications where accuracy is connected to safety and therefore most important, prioritizing speed by using faster, less deliberate decoding methods can directly compromise the reliability of the generated text. So what does this mean for people working with LLMs in healthcare? In my opinion, we need to shift our focus from just which model to use, to how we use it. This requires that we are recognizing that there is no one-size-fits-all solution and that decoding strategies must be evaluated and thought about for each specific medical use case. Second, treating specialized models not as a simple solution but as powerful tools that require even more careful tuning and testing to ensure their stability. And finally, deliberately prioritizing quality and consistency over raw inference speed in clinical applications (until we get models that can do both). with Oriana Presacan, Alireza Nik, Vajira Thambawita and Bogdan Ionescu Simula Research Laboratory Simula Metropolitan Center for Digital Engineering (SimulaMet) CAMPUS Research Institute University POLITEHNICA of Bucharest OsloMet – Oslo Metropolitan University #AI #LLM #Decoding #Healthcare

30

1 kommentar