Dr. Alessandra Sala

There is growing momentum in applying Generative AI (GenAI) to optimization and decision making. Our recent HBR article https://lnkd.in/erSugaKK highlights early examples from Microsoft and other companies. Yet, there is still a need for the OR and AI communities to think more holistically about how to maximize impact. Together with my colleagues Tinglong Dai, Ishai Menache, and Michelle Xiao Wu, I’m excited to share our new vision paper: “Democratizing Optimization with Generative AI.” We introduce the 4I framework — Insight, Interpretability, Interactivity, Improvisation — design principles that combine GenAI’s accessibility with optimization’s rigor. The aim: make optimization more intuitive, explainable, and adaptable, empowering frontline decision-makers to engage in rigorous choices. 👉 Read the paper here: https://lnkd.in/eM52Xe24

638

17 Comments

I spoke to Naga Munchetty on BBC Radio Five Live yesterday, alongside Kate M. Miltner, Ph.D about the government signing a strategic partnership with OpenAI to use AI in public services. Link in comments if you want to listen. Three things to remember: 1. ‘Productivity’ has to be defined by 'us' (in this case the Civil Service) rather than the tech companies. Research done by Institute for the Future of Work shows this means directing AI to understand and build national capabilities to respond to localised, regional priorities. Ref: Disruption Index 2. AI is not magic dust. We cannot add it in isolation to existing services and hope that productivity springs out. We must start with understanding our challenges and then thinking about the role that AI technologies play in addressing those challenges. That starts with the Civil Servants! 3. These technologies are here, and cannot be ignored, but that does not mean we must rush to a solution, nor does it mean we do nothing. The government must lead by example on how to put responsibility at the core of AI adoption to ensure these technologies serve to improve the society we live in. This requires tackling many questions (to name but a few!): use of data, data privacy and security, transparency of approach, understanding of what will be done with the productivity gains, sustainability, training and upskilling, workforce planning, ... There are many resources to help on this journey! Reach out if you're interested. Responsible Ai UK Institute for the Future of Work https://lnkd.in/eRah97Ne

85

4 Comments

What counts as plagiarism? AI-generated papers pose new risks https://lnkd.in/gW7RmwnA "And if more LLM-based tools are used to generate ideas, this could deepen the erosion of intellectual credit in science. Because LLMs work in part by remixing and interpolating the text they’re trained on, it would be natural for them to borrow from earlier work, says Parshin Shojaee, a computer scientist at the Virginia Tech Research Center — Arlington. The issue of ‘idea plagiarism’, although little discussed, is already a problem with human-authored papers, says Debora Weber-Wulff, a plagiarism researcher at the University of Applied Sciences, Berlin, and she expects that it will get worse with work created by AI. But, unlike the more familiar forms of plagiarism — involving copied or subtly rewritten sentences — it’s hard to prove the reuse of ideas, she says. That makes it difficult to see how to automate the task of checking for true novelty or originality, to match the pace at which AIs are going to be able to synthesize manuscripts."

52

🎉 New Publication Alert! I'm pleased to share that our paper, "A Review of Faithfulness Metrics for Hallucination Assessment in Large Language Models", has been published in the IEEE Journal of Selected Topics in Signal Processing. 🔍 This review explores how faithfulness is evaluated in LLM-generated outputs across summarization, question answering, and machine translation. We highlight the strengths and limitations of various metrics—including LLM-as-a-judge, embedding-based, and fact extraction approaches—and discuss promising strategies for mitigating hallucinations, such as retrieval-augmented generation and prompting frameworks. 📌 Faithfulness evaluation is critical for ensuring the reliability of LLMs, especially in safety-critical applications. We hope this work contributes to the development of more robust and trustworthy AI systems. 📖 Read the full paper here: https://lnkd.in/etFxezFH Grateful to my co-authors Ben Malin and Nikolaos Boulgouris for their collaboration and insights. #AI #LLMs #Faithfulness #HallucinationMitigation #IEEE #Research #SignalProcessing #MachineLearning #AcademicPublishing

38

3 Comments

🚀 Introducing #OntoLog: The Unified Intelligence Paradigm Experts and Entrepreneurs Can’t Ignore For too long, AI has been fragmented. We’ve had separate tools for #logic, #learning, and #reasoning—but nothing that brings them together seamlessly. Until now. I’m excited to share my latest Medium article, "OntoLog: The Unified Intelligence Paradigm That Experts and Entrepreneurs Can’t Ignore", where I introduce a new approach to AI that eliminates the trade-offs between #symbolic and #neural #systems. Why #OntoLog Matters Unified Intelligence: Combines #logic_programming, #probabilistic_reasoning, and #machine_learning in one language. Real-World Impact: Solves problems in healthcare, smart cities, and finance that current tools can’t handle. Rigorously Grounded: Built on formal foundations with proven soundness and completeness. Massive Market Opportunity: $120B+ addressable market by 2030. Who Should Care? 🔬 Experts: New research frontiers, publication opportunities, and collaboration. 💼 Entrepreneurs: First-mover advantage, scalable business model, and clear ROI. What You’ll Learn How #OntoLog handles uncertainty, ethics, and causality natively. Why it outperforms fragmented solutions by 10–100x. How it enables applications that are currently impossible. The Big Picture This isn’t just another tool—it’s a paradigm shift. OntoLog is redefining what’s possible in AI, and the time to get involved is now. 👉 Read the full article: OntoLog: The Unified Intelligence Paradigm, via the link below, please! https://lnkd.in/eDKZkfbZ I’d love to hear your thoughts! What problems would you solve with unified intelligence? #AI #MachineLearning #TechInnovation #FutureTech #Startup #Research #OntoLog #Knowledge_Representation #Ontology #Taxonomy

58

9 Comments