United We Care India’s Post

The fundamental nature of LLMs (underpinning #GenAI) is that hallucinations are just a part of how they work. For those in the world of #GovCon, just keep in mind that you need to "trust, but verify" what comes back from your prompts. Or the products you adopt. GenAI is excellent at getting you off the blank page, tasks like creating a 1st draft of your proposal section. Or tasks where "good enough" is fine. Just make sure you vet everything it comes back with. Use deterministic (logic) approaches when you need 100% accuracy & 100% repeatability. Tasks like document shreds, acronym pulls, accurate search etc.

So proud to announce that of our key research paper has gone live - LLMs Will Always Hallucinate, and We Need to Live With This has been published and gone live. As Large Language Models become more ubiquitous across domains, it becomes important to examine their inherent limitations critically. This work argues that hallucinations in language models are not just occasional errors but an inevitable feature of these systems. We demonstrate that hallucinations stem from the funda- mental mathematical and logical structure of LLMs. It is, therefore, impossible to eliminate them through architectural improvements, dataset enhancements, or fact- checking mechanisms. Our analysis draws on computational theory and Gödel’s First Incompleteness Theorem, which references the undecidability of problems like the Halting, Emptiness, and Acceptance Problems. We demonstrate that every stage of the LLM process—from training data compilation to fact retrieval, intent classification, and text generation—will have a non-zero probability of producing hallucinations. This work introduces the concept of "Structural Hallucinations" as an intrinsic nature of these systems. By establishing the mathematical certainty of hallucinations, we challenge the prevailing notion that they can be fully mitigated. This was led by our fantastic leadership team & cofounder Sourav Banerjee along with Ayushi Agarwal and Saloni Singla You can read more about it here - https://lnkd.in/gsybNN4W Ritu Mehrotra (She/Her) United We Care Deepanshu Nasa Sonakshi D. Torin Nicholas #Artificialintellegence #generativeAI #LLM #LargeLanguageModels #DeepTech #AI #startups #hallucination #Mentalhealth

AI policy alert: #SAPEA, a consortium of European academy networks, has published a comprehensive evidence review report examining how the European Commission can accelerate the responsible uptake of artificial intelligence (AI) in science to boost innovation, strengthen the EU's scientific position, and contribute to solving societal challenges. The report analyzes AI's impact on the scientific process, the skills and infrastructure needed for scientists, and policy considerations for ensuring timely and responsible AI adoption in European science. Key findings include: - The exponential growth of computational power required for advanced AI has created a divide between academia and industry in access to resources. - The USA, China, and the UK dominate globally in AI research, but the EU contributes significantly in social sciences and humanities. - AI offers opportunities to accelerate scientific discovery through literature mining, big data analysis, and automating research workflows. - Challenges include limited reproducibility, bias, ethical concerns over AI opacity, potential misuse, and threats to societal values. The report proposes policy options such as establishing an EU AI research facility, fostering best practices, developing researcher skills, evaluating scientific journals transparently, and monitoring societal threats from AI. A much more nuanced view by scientists compared with policymakers... a good reference for policy research / bench-marking (summary attached & detailed report link in comments ) #AI #ArtificialIntelligence #Science #Research #Innovation #EUPolicy #SAPEA #AcademicResearch #TechPolicy #ResponsibleAI #EthicalAI #DataScience #MachineLearning #ScientificAdvice

🏆 𝑪𝒆𝒍𝒆𝒃𝒓𝒂𝒕𝒊𝒏𝒈 𝑩𝒓𝒆𝒂𝒌𝒕𝒉𝒓𝒐𝒖𝒈𝒉𝒔 𝒊𝒏 𝑷𝒉𝒚𝒔𝒊𝒄𝒔 𝒂𝒏𝒅 𝑨𝑰! 🌟 This year’s #Nobel #Prize in #Physics has been awarded to two extraordinary scientists: British-Canadian 𝐆𝐞𝐨𝐟𝐟𝐫𝐞𝐲 𝐇𝐢𝐧𝐭𝐨𝐧 from the University of Toronto and American 𝐉𝐨𝐡𝐧 𝐇𝐨𝐩𝐟𝐢𝐞𝐥𝐝 from Princeton University, for their groundbreaking contributions that have revolutionized #machinelearning through 𝒂𝒓𝒕𝒊𝒇𝒊𝒄𝒊𝒂𝒍 𝒏𝒆𝒖𝒓𝒂𝒍 𝒏𝒆𝒕𝒘𝒐𝒓𝒌𝒔. Their research has been pivotal in laying the foundations for how modern AI systems learn and adapt. These advancements are not just academic; they’re opening a new area for IT and digital. And WEMAIK already made it! WEMAIK is the worldwide only digital platform built on a native neural network architecture Neural network offers the full flexibility, performance, and agility to generate complex enterprise applications in a fraction of time Neural Network combined with native Expert System offers the first AI able to 𝒓𝒆𝒑𝒍𝒊𝒄𝒂𝒕𝒆 𝒉𝒖𝒎𝒂𝒏 𝒕𝒉𝒊𝒏𝒌𝒊𝒏𝒈 to solve complex and dynamic problems WEMAIK Smart Digital Application Platform empowers a truly and always no-code approach for enterprise solutions ✅ 𝐀𝐥𝐥-𝐢𝐧-𝐨𝐧𝐞 𝐩𝐥𝐚𝐭𝐟𝐨𝐫𝐦, including all – native – functionalities to create any applications (without 3rd party technology) ✅ 𝐈𝐧𝐭𝐮𝐢𝐭𝐢𝐯𝐞 𝐛𝐮𝐬𝐢𝐧𝐞𝐬𝐬-𝐝𝐫𝐢𝐯𝐞𝐧, collaborative, and iterative way-of-working ✅ 𝐃𝐞𝐦𝐲𝐬𝐭𝐢𝐟𝐢𝐜𝐚𝐭𝐢𝐨𝐧 𝐨𝐟 𝐝𝐚𝐭𝐚 𝐜𝐨𝐦𝐩𝐥𝐞𝐱𝐢𝐭𝐲 whatever format/source, IT integration and other traditional IT bottlenecks ✅ 𝐀𝐛𝐢𝐥𝐢𝐭𝐲 𝐭𝐨 𝐬𝐜𝐚𝐥𝐞, adjust, enhance at any time and on one-click (real time) ✅ 𝐄𝐧𝐫𝐢𝐜𝐡 𝐚𝐧𝐝 𝐜𝐨𝐧𝐭𝐞𝐱𝐭𝐮𝐚𝐥𝐢𝐳𝐞 𝐞𝐱𝐢𝐬𝐭𝐢𝐧𝐠 𝐌𝐋 𝐦𝐨𝐝𝐞𝐥𝐬 with “human AI” for better decision making, recommendation, simulation, etc. It’s inspiring to see how the scientific achievements of visionaries like Hinton and Hopfield are enabling companies like WeMaik to push the boundaries of innovation. 💡🚀 Let’s continue turning groundbreaking research into game-changing solutions! #NobelPrize #AI #Innovation #MachineLearning #WeMaik #DigitalTransformation #NoCode #NeuralNetwork #AppliedAI Xavier Peny, Swiss InsurTech Hub, Brijesh Luthra, Anna Nevano, Sharanjit Kaur, Pawel Konieczniak, David Sebastio, Alessandro Paoli, Massimo Michaud, Roberta Profeta, Indumathi Chitharanjan, Babak Hodjat

Peter Norvig's quote emphasizes the potential of AI to be applied to a wide range of tasks and domains, as long as they can be mathematically described. Norvig's statement highlights the potential for AI to: - Be applied to various fields and industries, such as science, engineering, and economics - Solve complex problems and optimize outcomes using mathematical models - Continue to expand its capabilities as new mathematical descriptions are developed With the understanding that AI's potential to be applied to mathematically described domains, Norvig's quote encourages us to explore new applications and developments in AI research.

🤖 Exciting breakthrough in AI evaluation! Our latest article explores how researchers are revolutionizing the way we test AI models with the Dynamic Intelligence Assessment (DIA) framework. Ever wondered how we can truly measure an AI's problem-solving abilities? Traditional benchmarks might not be telling us the whole story! 🤔 A groundbreaking collaboration between Technology Innovation Institute, University of Oslo, Eötvös Loránd University, Université du 8 mai 1945 Guelma, The University of Manchester, and Khalifa University has developed a dynamic approach that's changing the game in AI evaluation. [Link to full article in comments ⬇️] Read our in-depth analysis on EngiSphere to learn: 📉 Why traditional benchmarks fall short 📊 How DIA provides a more complete picture 🚀 What this means for the future of AI reliability 🎯 Question for discussion: How do you think dynamic testing methods like DIA could impact the future of AI development in your industry? 🌟 Follow EngiSphere for daily engineering insights that keep you at the forefront of innovation! Remember: being cultivated means knowing something about everything. Share this knowledge with your network! Special thanks to researchers: Norbert Tihanyi, PhD, Tamas Bisztray, PhD, Richard A. Dubniczky, Rebeka Toth, Bertalan Borsos, Bilel Cherif, Ridhi Jain, Lajos Muzsai, Mohamed Amine Ferrag, PhD, Ryan Marinelli, Lucas C. Cordeiro, and Merouane Debbah for this impressive work! 👏 #ArtificialIntelligence #Engineering #Innovation #AIResearch #TechInnovation #DataScience #MachineLearning #STEM #EngineeringCommunity #ResearchAndDevelopment #AIEvaluation #TechNews #Innovation #FutureOfAI #EngiSphere

Are we approaching AI with PhD-level intelligence? 🤔 The comparison of Terence Tao (a famous mathematician) of AI to a "𝘮𝘦𝘥𝘪𝘰𝘤𝘳𝘦, 𝘣𝘶𝘵 𝘯𝘰𝘵 𝘤𝘰𝘮𝘱𝘭𝘦𝘵𝘦𝘭𝘺 𝘪𝘯𝘤𝘰𝘮𝘱𝘦𝘵𝘦𝘯𝘵" graduate student in a Mastodon Post went viral on social media, with 𝐬𝐨𝐦𝐞 𝐢𝐧𝐭𝐞𝐫𝐩𝐫𝐞𝐭𝐢𝐧𝐠 𝐭𝐡𝐢𝐬 𝐚𝐬 𝐞𝐯𝐢𝐝𝐞𝐧𝐜𝐞 𝐨𝐟 𝐀𝐈'𝐬 𝐢𝐦𝐦𝐢𝐧𝐞𝐧𝐭 𝐏𝐡𝐃-𝐥𝐞𝐯𝐞𝐥 𝐜𝐚𝐩𝐚𝐛𝐢𝐥𝐢𝐭𝐢𝐞𝐬. However, Tao clarified that his statement was limited to AI's ability to assist with (tedious) computational aspects of research, not a comprehensive equivalence to graduate-level expertise. In a recent article in The Atlantic, Tao elaborates further on AI's role in mathematics.  𝐇𝐞 𝐞𝐦𝐩𝐡𝐚𝐬𝐢𝐳𝐞𝐬 𝐯𝐢𝐞𝐰𝐢𝐧𝐠 𝐀𝐈 𝐚𝐬 𝐚 𝐭𝐨𝐨𝐥 𝐭𝐨 𝐞𝐧𝐡𝐚𝐧𝐜𝐞 𝐡𝐮𝐦𝐚𝐧 𝐜𝐫𝐞𝐚𝐭𝐢𝐯𝐢𝐭𝐲 𝐫𝐚𝐭𝐡𝐞𝐫 𝐭𝐡𝐚𝐧 𝐚 𝐫𝐞𝐩𝐥𝐚𝐜𝐞𝐦𝐞𝐧𝐭 𝐟𝐨𝐫 𝐡𝐮𝐦𝐚𝐧 𝐢𝐧𝐭𝐞𝐥𝐥𝐞𝐜𝐭. Tao advocates for developing AI systems that augment our existing capabilities instead of attempting to supplant human strengths. This perspective invites us to rethink how we frame the progress of AI and its integration into complex fields such as mathematics and scientific research. Computing a solution is only a small part of the research process: generating innovative ideas, framing them in relevant questions and making the most of the results are still tasks that rely on human intelligence, creativity and intuition. Let's avoid phrases like: "AI will cure diseases" - it is like saying "computers sent us to the moon" - and 𝐫𝐞𝐜𝐨𝐠𝐧𝐢𝐳𝐞 𝐭𝐡𝐞 𝐚𝐦𝐚𝐳𝐢𝐧𝐠 𝐡𝐮𝐦𝐚𝐧 𝐰𝐨𝐫𝐤 𝐛𝐞𝐡𝐢𝐧𝐝 𝐞𝐯𝐞𝐫𝐲 𝐬𝐜𝐢𝐞𝐧𝐭𝐢𝐟𝐢𝐜 𝐝𝐢𝐬𝐜𝐨𝐯𝐞𝐫𝐲 💪 🗯 #AI can be a very useful tool, but #HumansAreAmazing BFH Humane Digital Transformation

This is the last part of the entropy thread, bringing it home to AI. In the rapidly evolving field of AI, while we focus on processing power, data availability, and algorithm sophistication, we must consider a fundamental aspect of our universe: the information force. Just as energy and mass are essential to physics, information is emerging as equally important. The Second Law of Infodynamics, which describes how information entropy tends to decrease or remain constant, has profound implications for AI. The universe naturally evolves towards states that optimize information persistence and work. Gravitational collapse, quantum coherence, and even DNA mutations demonstrate nature’s efficiency in processing information. By aligning AI systems with these natural tendencies of information in the universe, we may unlock new potentials for efficiency, organization, and problem-solving. The information force is not just theoretical – it’s a practical consideration shaping the future of AI. Read more at - https://lnkd.in/eqrhSzNj #InformationEntropy #AI #DataScience #Innovation #Tech #MachineLearning

What AI can do for science? “AI in Science” - a policy brief from the European Commission answers this question, but also contains a lot of other interesting content: AI is changing the scientific landscape at every stage of research, assisting researchers, making predictions (learning from trends, correlations, patterns), automating tasks (literature review processing, summarizing papers, mining historic literature archives, using self-driving labs), processing large-scale data (also improving its quality), as well as automating reports, visualizations or summaries. According to the report: “AI is likely to be the main driver of discovery and innovation in the future...” Through Horizon Europe and the Digital Europe program the Commission is investing €1 billion per year in AI. The EU wants to create the framework and conditions for excellent and trustworthy AI. It seeks not only to adapt available AI tools for science, but to build AI tools dedicated to the needs of researchers. The report also emphasizes the need to train non-AI scientists in computational and data-handling skills to skillfully integrate AI into their work (so-called “hybrid skills”). Also key is data access (open science, open data, FAIR, reuse of data, interoperability), access to AI tools and computer resources, as well as funding for data annotation/labeling. AI is not meant to replace researchers, but to amplify human abilities. "The expertise, judgement, creativity, and critical thinking of human scientists remain crucial to maintaining the integrity and reliability of scientific work." Public-private collaboration in AI-powered R&D is necessary, as is academic-industry collaboration. “In order to remain competitive at EU and global level, university and research departments need to prepare their organisational processes, staff and researchers for more intense use of AI in science and increased cross-displinary work”.

The universe is vast, and the development of life as we know it has taken billions of years. In contrast, the development of intelligence has taken very little time. But that's about to change. With the introduction of super-intelligence, we're poised to witness an unprecedented scale of advancement. Artificial Intelligence (AI) is a topic that few people fully understand, and we often avoid discussing it. However, we need to start thinking about what we want the future of AI to look like because our desires will shape its outcome. It's essential to remember that technology is limited by the laws of physics, just as the development of human life. But that doesn't mean we can't make a difference. So let's start thinking about what we want the aftermath of AI to be.