Nicholas Clarke’s Post

This paper runs generative agent simulations of 1,000 people 🤔 Agents have been developed to simulate the attitudes and behaviors of 1,000 real individuals. 💁 This is accomplished by applying large language models to qualitative interviews and measuring how well these agents replicate the attitudes and behaviors of the individuals they represent. 🤖 The generative agents were able to predict participants' responses on unseen surveys (e.g. on the General Social Survey with 85% accuracy, comparable to participants reproducing their own responses two weeks later). They also perform similarly in predicting personality traits and outcomes in experimental replications. 📊 The work provides a foundation for new tools to research individual and collective behavior. It's basically a laboratory for testing broad sets of interventions and theories. These simulations could pilot interventions and help expand understanding across various domains, including economics, sociology, and political science. Source: https://lnkd.in/e5saYXbN —————————————————— Follow Generative AI Central for daily insights into the most relevant AI news, guides, and research! 🚀

This paper pertains to AI agents simulating human society. It features 25 agents in a small virtual town like the Sims. Researchers (from Google Deepmind and Stanford) observed a curious fact: Instructing a single agent to throw a Valentine day's party, the agents autonomously spread the invitation, made new acquaintances, and even asked each other on date. In short, they behave like humans, like a human society. _ Could agentic AI serves as society simulacra for social studies and simulations? Many fields such as Politics, Pandemics, Sociology, Military, Macroeconomics can benefit from such large scale simulations. What an awesome time to leave in!

I'm excited to announce the release of my first published book “Philosophy of Artificial Intelligence and Images of the Science of the Future”, co-authored with Prof. Ilham Mammadzada, Prof. Tahir Makhamatov and Prof. Arif Buniyatov. Publisher: Institute of Philosophy and Sociology of ANAS. Artificial Intelligence and digitalization of almost all spheres of life have led to a change in the consciousness of a scientist and research methods and, as a consequence, to the birth of new theories and concepts that can expand our understanding of the world order. The contribution of AI to science is very determined by raising new philosophical questions, ideas, and categories as well as by the formation of a new sound for old, perhaps long-forgotten questions. The essence of this approach is the constant updating of scientific practices through the identification of general patterns and connections, the integration and synthesis of ideas from various disciplines. The interdisciplinary nature of the materials presented in the work allows us to cover not only the philosophy of AI itself but also problems of consciousness and cognition, subjectivity and reality. 

🚀 𝐁𝐡𝐚𝐫𝐭ṛ𝐡𝐚𝐫𝐢'𝐬 स्फोटा 𝐨𝐫 **𝐄𝐮𝐫𝐞𝐤𝐚!! 𝐅𝐥𝐚𝐬𝐡** 𝐢𝐧 𝐋𝐋𝐌 𝐔𝐧𝐝𝐞𝐫𝐬𝐭𝐚𝐧𝐝𝐢𝐧𝐠𝐬 🧠 🔍 In my talk at Matrix Global Summit I delved into Epistemology (Philosophy of Meaning) and Sphota. This concept offers a fascinating angle to the debate on whether LLMs are just  𝐣𝐮𝐬𝐭 "𝐬𝐭𝐨𝐜𝐡𝐚𝐬𝐭𝐢𝐜 𝐩𝐚𝐫𝐫𝐨𝐭𝐬" 𝐨𝐫 𝐭𝐫𝐮𝐥𝐲 𝐮𝐧𝐝𝐞𝐫𝐬𝐭𝐚𝐧𝐝 𝐦𝐞𝐚𝐧𝐢𝐧𝐠. 🤔 For context, let's reference Anthropic's presentation on 𝐦𝐨𝐧𝐨𝐬𝐞𝐦𝐚𝐧𝐭𝐢𝐜𝐢𝐭𝐲. Neurologists have mapped human thinking to specific brain regions and neuron clusters. Similarly, geneticists have linked diseases to specific genes. Both fields reveal layers of abstraction. 🧬🧠 In LLM embeddings or pre-training, we reduce the corpus to a smaller vocabulary (26-50K words) and map them to a semantic space. These tokens, which may be root words ignoring tense and gender, are represented in vectors. The activation of one, two, or three cells of the vector represents not just the token but also maps to a concept in the semantic space. 🌐📚 The Anthropic exploration has assigned numbers to different combinations of activations, mapping concepts to higher layers of abstraction.  **𝘉𝘩𝘢𝘳𝘵ṛ𝘩𝘢𝘳𝘪 𝘮𝘢𝘪𝘯𝘵𝘢𝘪𝘯𝘦𝘥 𝘵𝘩𝘢𝘵 𝘵𝘩𝘦 𝘦𝘯𝘵𝘪𝘳𝘦 𝘴𝘦𝘯𝘵𝘦𝘯𝘤𝘦 𝘪𝘴 𝘢 𝘮𝘦𝘢𝘯𝘪𝘯𝘨 𝘶𝘯𝘪𝘵, 𝘯𝘰𝘵 𝘵𝘩𝘦 𝘪𝘯𝘥𝘪𝘷𝘪𝘥𝘶𝘢𝘭 𝘸𝘰𝘳𝘥𝘴.** 𝘍𝘰𝘳 𝘪𝘯𝘴𝘵𝘢𝘯𝘤𝘦, "𝐑𝐞𝐝 𝐒𝐞𝐚" 𝘤𝘰𝘶𝘭𝘥 𝘳𝘦𝘧𝘦𝘳 𝘵𝘰 𝘢 𝘴𝘦𝘢 𝘳𝘦𝘥𝘥𝘦𝘯𝘦𝘥 𝘣𝘺 𝘤𝘰𝘭𝘰𝘳, 𝘢 𝘨𝘦𝘰𝘨𝘳𝘢𝘱𝘩𝘪𝘤𝘢𝘭 𝘭𝘰𝘤𝘢𝘵𝘪𝘰𝘯, 𝘰𝘳 𝘢 𝘨𝘦𝘰𝘱𝘰𝘭𝘪𝘵𝘪𝘤𝘢𝘭 𝘵𝘩𝘦𝘮𝘦 𝘪𝘯 𝘵𝘩𝘦 𝘐𝘴𝘳𝘢𝘦𝘭-𝘎𝘢𝘻𝘢 𝘤𝘰𝘯𝘧𝘭𝘪𝘤𝘵. 𝘈 𝘴𝘰𝘱𝘩𝘪𝘴𝘵𝘪𝘤𝘢𝘵𝘦𝘥 𝘓𝘓𝘔 𝘸𝘪𝘭𝘭 𝘮𝘢𝘱 𝘵𝘩𝘪𝘴 𝘤𝘰𝘳𝘳𝘦𝘤𝘵𝘭𝘺 𝘵𝘰 𝘥𝘪𝘧𝘧𝘦𝘳𝘦𝘯𝘵 𝘴𝘦𝘮𝘢𝘯𝘵𝘪𝘤 𝘴𝘱𝘢𝘤𝘦𝘴. 🗺️🌊⚔️ Idioms as tokens too... The Anthropic exploration has assigned numbers to different combinations of activations, mapping concepts to higher layers of abstraction. 𝐓𝐡𝐢𝐬 𝐢𝐧𝐧𝐞𝐫 𝐜𝐨𝐧𝐟𝐥𝐢𝐜𝐭 𝐦𝐚𝐩 𝐬𝐡𝐨𝐰𝐜𝐚𝐬𝐞𝐬 𝐜𝐨𝐧𝐜𝐞𝐩𝐭𝐬 𝐭𝐡𝐚𝐭 𝐰𝐨𝐮𝐥𝐝 𝐢𝐦𝐩𝐫𝐞𝐬𝐬 𝐩𝐡𝐢𝐥𝐨𝐬𝐨𝐩𝐡𝐞𝐫𝐬, 𝐩𝐬𝐲𝐜𝐡𝐨𝐥𝐨𝐠𝐢𝐬𝐭𝐬, 𝐬𝐨𝐜𝐢𝐨𝐥𝐨𝐠𝐢𝐬𝐭𝐬, 𝐚𝐧𝐝 𝐩𝐨𝐥𝐢𝐭𝐢𝐜𝐚𝐥 𝐬𝐜𝐢𝐞𝐧𝐭𝐢𝐬𝐭𝐬 𝐚𝐥𝐢𝐤𝐞. 🧩🔍 𝘛𝘩𝘪𝘴 𝘵𝘪𝘦𝘴 𝘪𝘯𝘵𝘰 𝘉𝘩𝘢𝘳𝘵ṛ𝘩𝘢𝘳𝘪'𝘴 𝘥𝘦𝘦𝘱𝘦𝘳 𝘤𝘭𝘢𝘪𝘮: 𝘸𝘦 𝘯𝘦𝘷𝘦𝘳 𝘪𝘯𝘷𝘦𝘯𝘵 𝘢𝘯𝘺𝘵𝘩𝘪𝘯𝘨 𝘣𝘶𝘵 𝘥𝘪𝘴𝘤𝘰𝘷𝘦𝘳. 𝘛𝘩𝘦 𝘓𝘢𝘵𝘪𝘯 𝘵𝘦𝘳𝘮 **𝘳𝘦𝘱𝘳𝘢𝘦𝘴𝘦𝘯𝘵𝘢𝘳𝘦** 𝘪𝘮𝘱𝘭𝘪𝘦𝘴 𝘢 𝘤𝘰𝘱𝘺 𝘰𝘳 𝘢 𝘴𝘪𝘨𝘯𝘪𝘧𝘪𝘦𝘳 𝘰𝘧 𝘢𝘯 𝘦𝘹𝘪𝘴𝘵𝘪𝘯𝘨 𝘢𝘳𝘵𝘪𝘧𝘢𝘤𝘵. 𝘛𝘩𝘢𝘵 𝘪𝘴 𝘵𝘩𝘦 𝘧𝘭𝘢𝘴𝘩. 𝘐𝘯 𝘵𝘩𝘪𝘴 𝘤𝘢𝘴𝘦, 𝘴𝘺𝘮𝘣𝘰𝘭𝘪𝘤 𝘱𝘳𝘰𝘤𝘦𝘴𝘴𝘪𝘯𝘨 𝘮𝘢𝘤𝘩𝘪𝘯𝘦𝘴 𝘭𝘪𝘬𝘦 𝘓𝘓𝘔𝘴 𝘥𝘰 𝘶𝘯𝘥𝘦𝘳𝘴𝘵𝘢𝘯𝘥. 💡✨ 📚 For further reading, check out comments Matrix Forum #LLM #AI #Epistemology #Sphota #Understanding #ArtificialIntelligence #Philosophy #Neuroscience #SemanticSpace #Innovation #claude #chatGPT

🚨 New publication just dropped!🚨 Congratulations to Tom Chan, PhD Researcher from the ANU School of Cybernetics, and co-authors. ✨ This new publication looks at #Artificial Intelligence (AI), #Public Policy, and the #Sociology of Expectations through examining how AI is (or isn't) actually being developed and used to address #sustainable policy problems.✨ You can read this new research here: https://lnkd.in/eQbNR3Da

🤔 Are Traditional Models Missing the Mark in Social Science? 🤔 As social phenomena become more complex, can linear models truly capture the intricate relationships at play? The latest chapter explores how machine learning (ML) is revolutionizing social science research by effectively managing nonlinear interactions. From decision trees to neural networks, this chapter reveals how advanced techniques uncover hidden patterns in human behavior and societal dynamics. 📊✨ It also delves into the practical challenges and ethical considerations essential for successful ML application in research. It’s time to embrace a future where data-driven insights inform psychology, sociology, and political science! 👉 https://lnkd.in/gbZhGBjk #MachineLearning #SocialScience #DataAnalysis #NonlinearModels #ResearchInnovation #PredictiveAnalytics

#CallforPapers| A special issue of #EasternEuropeanHolocaustStudies is announced on The Quantitative Turn in Holocaust Research. This special issue explores the growing role of quantitative methods, ranging from statistical techniques to natural language processing to deep learning and data visualizations in Holocaust research. Scholars from diverse fields, including history, sociology, demography, economics, and digital humanities, are invited to submit abstracts that showcase innovative research, methodologies, and case studies that illuminate new perspectives on the Holocaust through these advanced techniques. Submit your proposal by December 6, 2024. More about the issue and the calendar of submission:https://lnkd.in/d8qEc7Fm

🔍 Exciting News Alert! 🚀 Cutting-edge research in Denmark delves into the transformative potential of Artificial Intelligence (AI) on knowledge production. Two groundbreaking studies spearheaded by experts in computer science and sociology of science at the University of Copenhagen are set to unveil the impact of generative AI on scientific practices. With funding from Villum Fonden and the European Research Council, these projects aim to map AI's influence on science, analyze its current and future impacts, and explore AI's role in shaping scientific knowledge. Led by Professor Roberta Sinatra, the research endeavors to compare human-generated knowledge with AI outputs, shedding light on the unexpected consequences of AI-infused science. Stay tuned for groundbreaking insights into the evolving landscape of scientific discovery! 🌟 #AIResearch #ScienceInnovation #KnowledgeProduction https://lnkd.in/dnmvS7EF Jan Petter Myklebust University World News

𝗕𝗲𝘁𝘄𝗲𝗲𝗻 𝗘𝗺𝗽𝗶𝗿𝗶𝗰𝗮𝗹, 𝗡𝗼𝗻-𝗘𝗺𝗽𝗶𝗿𝗶𝗰𝗮𝗹 𝗮𝗻𝗱 𝗪𝗵𝗮𝘁 𝗴𝗼𝘁 𝗯𝘂𝗿𝗻𝘁!! There have been innumerable discussions about "Psychology as a Science;" But what exactly is "Science as a whole?" Let us break it down: Thomas Kuhn in his book, The Structure of Scientific Revolution" explains how scientific progress happens: Normal Science : Scientists work within an accepted set of ideas. Anomalies : Sometimes, results don’t fit the current ideas. Crisis: Many anomalies lead to doubts about the current paradigm. Paradigm Shift : New set of ideas emerge to better explain the anomalies. Scientific Revolution: The shift from old ideas to new ones is a revolution in science. Empirical studies in psychology follow a similar pattern where researchers work within established theories - new findings that don’t fit these theories appear (anomalies) - accumulation of conflicting results challenges the prevailing theories (crisis) - new theories start offering better explanations (paradigm shift) - acceptance of new theories over old ones marks a significant change in the field (revolution). 📌 Now does that mean everything that does not fit into "empirical" data is automatically a "pseudoscience?" Let us understand : Between the polarized concepts of empirical v/s pseudoscience, there is something called as "𝗹𝗲𝗴𝗶𝘁𝗶𝗺𝗮𝘁𝗲 𝗻𝗼𝗻-𝗲𝗺𝗽𝗶𝗿𝗶𝗰𝗮𝗹 𝗱𝗮𝘁𝗮" that includes theories and qualitative insights that still hold value. Pseudoscience vs. Non-Empirical Knowledge While pseudoscience claims look scientific but lack evidence or cannot be tested, legitimate non-empirical knowledge involves centuries of observational studies and qualitative studies that don’t make false claims, for example, Vedanta, Nyaya, Samkhya, Mimasa along with Bhagavatha, Tantra, etc. Unfortunately, one would not know of very many studies that may or may not have been quantitative, as they were burnt to ashes during invasions. 📌 But what about practical implementation in therapeutic setting in today's day & age? Here, it becomes essential to learn the difference between "applicable science" and "science that needs more digging." Simply put, would we have dared to take vaccinations if they were not tried and tested? Would we have dared to try out a lot of things in life if it was not for the sense of believability that seeing others do it with no or no catastrophic consequence brings us? Empirical studies definitely bring in structure, reliability, safety, predictability and efficacy when it comes to treatment. It is a matter of responsibility when another life is involved. It is safer to rely on something concrete than introducing something based on one's own sentiment. Having said that, mental health is not limited to something clinical & pathological always, mental health is also "psycho-socio-cultural" that shapes and creates "mindset." This is where cultural learnings plays a major role in creating "mental fitness."