From Trustworthy AI to Regulation and Diplomacy

Professor Barry O'Sullivan

Bio: Professor O'Sullivan is an award-winning academic with more than 25 years experience working in artificial intelligence. He is Professor of Constraint Programming at the School of Computer Science & IT at University College Cork and a member of its Governing Body. He is founding Director of both the Insight Centre for Data Analytics at UCC and Director of the SFI Centre for Research Training in AI. In July 2018 Barry was appointed Vice Chair of the European Commission High-Level Expert Group on AI. He is a Fellow and a past President of the European AI Association. He is also a Fellow and a member of the Executive Council of the Association for the Advancement of Artificial Intelligence. He is a member of the Royal Irish Academy. He chairs the Advisory Board of the GRACE project at Europol, and advises the institute at KULeuven (Belgium) and the Computational Sustainability Network, a network of universities in the USA. He is Senior Advisor for Technology Policy at INHR in Geneva. In 2019 Professor O’Sullivan was appointed by Ireland’s Minister for Health to the Health Research Consent Declaration Committee. In 2020 he was appointed Chair of the Oversight Board of Health Data Research UK (North), led by the University of Liverpool. In 2021 he was, again, appointed by the Minister for Health as Chair of the National Research Ethics Committee for Medical Devices. In 2022 he was appointed by the Minister for Trade Promotion, Digital & Company Regulation to the Enterprise Digital Advisory Forum. Show More

Abstract: There has been an increasing interest in the role of ethics in artificial intelligence and in the notion of Trustworthy AI, in particular. The European Commission has invested significant effort in setting out its strategy and its vision for the importance of delivering Trustworthy AI, including the proposal of a new regulatory instrument, the AI Act. In this talk I will give an overview of the policy context in relation to Trustworthy AI. I will give an overview of the current European Union’s AI Act and its origins. While most of the discussion around AI tends to be in the civilian domain, I will also give a brief overview of the diplomacy of AI. Show More

The nitty gritty of Principal Component Analysis

Professor Aedin Culhane

Bio: Aedín Culhane is a Professor of Cancer Genomics and Director of the Limerick Digital Cancer Research Center in the University of Limerick, School of Medicine. She is a computational biologist with expertise in multimodal data integration, statistical genomics and clinical bioinformatics. She has over 20 years’ experience in cancer genomics, with 15 of these in the Data Science Department at Dana-Farber Cancer Institute and Harvard University. She is a member of the Human Cell Atlas project that will describe the molecular profile of every human cell. This is an enormous undertaking as its estimated there are 37.2 trillion cells in the human body. Aedin is a leader in the Bioconductor community, a global open source, open development platform written in R for analysis of genomics data. She leads the eHealth-Hub for Cancer for federated digital health research in cancer. Her recent research focuses on development of matrix factorization approaches, machine learning and statistical algorithms for integrative analysis of multi-modal high dimensional sparse molecular data. Show More

Abstract: Principal component analysis (PCA), is a popular matrix factorization method that is often the first step in machine learning pipelines. It is a relatively fast dimension reduction method that can scale to large datasets. I will review PCA, the relationship between PCA and singular value decomposition, the difference between PCA of a correlation and covariance matrix, the impact of scaling, log-transforming, and standardization, and how to recognize a horseshoe or arch effect in a PCA. I will describe extensions to PCA and matrix factorization approaches for the integration of multi modal data. I will discuss why pre-processing or weighting datasets within a joint or multi-dataset decomposition should be considered. These discussions will be applied to high dimensional sparse single cell molecular data in cancer, where datasets are typically thousands of rows by tens of thousands to millions of columns. Show More

Lindsey Kostas

Bio: Lindsey is a Senior Staff Machine Learning Researcher. She joined a nascent ML R&D team at Qualcomm in 2018 and since that time she has led multiple projects in ML-based CAD/EDA which have impacted global SoC design process for teams across the globe leading to significant savings in time-to-market, compute and NRE cost. She holds four pending patents related to this work and consults on a variety of ML-driven initiatives across the company in application ranging from digital and analog design to 5G to licensing. In 2021, she was honored by the Global Semiconductor Association (GSA) as the inaugural Female Up-And-Comer for her exceptional contributions toward the development, innovation, growth, and success in the semiconductor industry. Prior to joining Qualcomm, Lindsey was a 4-year scholarship athlete at Stanford University where she won two tennis national team championships and was honored as an Elite 89 Award Finalist. After graduating with distinction in Economics, she obtained her master’s degree in Computer Science with an emphasis in Artificial Intelligence from Stanford University. While in the master’s program Lindsey was a teaching assistant and a research associate for Chris Re and Jure Leskovic with an emphasis in deep representation learning. Show More

Abstract: Since its inception, Qualcomm has led mobile SoC integration with the continuous production of next-generation IP. More recently this leadership in IP and integration has scaled beyond mobile to support everything from automotives to smart cities and home to personal devices with high performance low power on-device intelligence for everything wireless. This talk will highlight the challenges presented in the face of increasingly diverse and complex product applications and customer specifications; how state-of-art machine learning is and can be employed to address these challenges; and how one can prepare themselves to make an impact in the space of ML-based SOC Design. Show More