Computer Science and Engineering | University at Buffalo’s Post

AI and data science is bringing significant social, economic, and health benefits to the world. Using AI and data science techniques, digital machines can analyze and learn from big datasets, and discover more efficient ways to do complex tasks and make intelligent decisions with much higher accuracy and speed than human beings. Enroll to MSc Applied AI and Data Science at Solent university for this Jan 2025!! For more details: https://www.solent.ac.uk/ #studyinuk #studyaboard #unitedkingdom #Datascience #solentuniversity

My finalized slides & detailed speaker notes from yesterday's talk on AI countergovernance at University of Essex are now available: https://lnkd.in/eSu-xEFD If you missed the talk or would like a quick recap, you can read through the slides and notes. Huge thanks to Phoebe Moore for hosting and to all those who attended online! I had a wonderful time speaking with our small and engaged group. Abstract: The governance of artificial intelligence has recently become a strategic imperative for state and industry actors. However, vulnerable workers and communities are often pushed to the margins of top-down AI governance initiatives led by state and industry power. How can we oppose AI governance that fails to serve our interests? Highlighting lessons learned from backlashes against Google's Project Maven, Sidewalk Toronto, Canada's Artificial Intelligence & Data Act, and the Writers Guild of America labor strike, this talk will provide guidance and tools for building counterpower against top-down AI governance initiatives that exclude vulnerable workers and communities.

I remember picking this book up back in 2016ish... and not fully comprehending what these experts were saying. I'm re-reading many of these essays now, and it's incredible to understand how spot-on they were, while we're living and interacting with these LLMs in 2024. Here's a paraphrased summary of one such essay by Jon Kleinberg, Professor of Computer Science at Cornell and Sendhil Mullainathan, Professor of economics at Harvard -- called "We Built Them, But We Don't Understand Them" "As algorithmic technology advances, it becomes simultaneously more intelligent and less comprehensible. Although we create these algorithms and understand their basic functions, the complex behaviors they exhibit can surpass our understanding due to the billions of operations they perform. This irony poses a challenge, as our human domain knowledge may not reliably predict which variables will enhance algorithm performance. Consequently, we must reconsider our traditional views on comprehensibility. It might be sufficient to learn how to effectively interact with these systems, trusting and leveraging their capabilities in ways that surpass our own, even if we cannot grasp every detail of their processes." #ai #artificialintelligence #machines #thinkingmachines #llms #technology #future #comprehensibility #machineintelligence #innovation #thoughtleadership

Questioning whether generative AI will be a benevolent or malevolent force in society? We had an interesting conversation at Digital Data Design (D^3) Institute at Harvard. Our Principal Investigators in the Data Science Operations lab, Iavor Bojinov and Edward McFowland III, chatted with Bratin Saha of Amazon. I'll give you sneak peek to the convo. Next time, join us in Cotting Hall! #generativeai #ai #datascience

Thrilled to have attended an enlightening lecture on the intersection of AI, data, and the economy, led by Mike Keoghan from the National Office of Statistics. As a final year student at NTU, this experience was a perfect blend of academia and practical insights. It's given me fresh perspectives for my ongoing journey and research in AI. Exciting times ahead! 🌟💡#ai #data #economy #ntu

🚀The Intersection of Science and Machine Learning 📗 Throughout history, humans have looked at the sky, stars, sun, and the natural world, recognized #patterns in them that enabled the creation of agriculture, calendar, medicines etc. But they did not ask about the origin of phenomena and rather attributed them to various gods. 🔬The advent of the scientific method, largely popularized by the works of Ibn al-Haytham on optics, marked a pivotal shift in our approach to understanding the world. But what exactly is the scientific method? At its core, it consists of formulating #hypotheses, deriving logical #predictions from them, and conducting #experiments based on these predictions to ascertain the validity of the original hypotheses (see the below image from Wikipedia). 🔑 While the scientific method not only aims to #predict outcomes but also seeks to #explain the underlying mechanisms, the same cannot be said for Machine Learning (ML) methods. ML predominantly focuses on black-box #prediction and pattern recognition, lacking the crucial element of explaining "how" and "why" things work. Nevertheless, ML algorithms serve as invaluable tools for scientists, enabling them to process vast datasets and uncover patterns that accelerate hypothesis testing. #science #ai #algorithms #machinelearning #history

Our Opening Keynote - Luis Seco - Professor at the Department of Mathematics at the University of Toronto joined us at our first-ever conference in the Middle East - DSC MENA! 🙌 His talk was on "Data As The New Commodity: AI, Sustainability and Financial Innovation" Thank you once more for being part of this conference, and we look forward to seeing you again on stage! #ai #datascience #ml #cairo #egypt #datadriven #generativeai #bigdata

Thought leaders like Nitesh Chawla are making it possible to transform bold new concepts into practical applications that we can use in the real world. Great going, Nitesh Chawla

How can technology advance the common good? Check out this interview with faculty fellow Nitesh Chawla!

📢 Exciting News! 📢 Our latest paper "Beyond the Norms: Detecting Prediction Errors in Regression Models" has been selected as a spotlight-designated paper at ICML 2024 (3.5% acceptance rate on 10,000 papers). 🌟 Joint work with CNRS - Centre national de la recherche scientifique, CentraleSupélec - Paris-Saclay University Research Center, New York University, Technische Universität Wien, Mila - Institut québécois d'intelligence artificielle and ILLS. Machine learning models can yield unreliable predictions due to their inability to generalize to out-of-distribution data, vulnerability to adversarial attacks, and other anomalies stemming from inherent model uncertainty. Tackling these issues necessitates robust uncertainty quantification, particularly in regression tasks where a single output often fails to adequately reflect prediction reliability. In this work, we introduce a data-driven framework to detect prediction errors in regression tasks by leveraging Rao's measure of statistical diversity. This innovative approach aims to enhance the robustness and accuracy of ML models. For further details, check: https://lnkd.in/efbKwJAm Looking forward to seeing you at ICML 2024! #MachineLearning #Research #AI #UncertaintyQuantification #RegressionAnalysis #DataScience