Artificial Intelligence for Mathematical Discovery

Join via the online stream (also to view the archived talk).

Abstract:

Progress in mathematics often involves observing a large number of well distributed examples, postulating conjectures and rigorously proving them. In a physicist’s language, this is a “top-down” approach and many spectacular results in both these disciplines have been discovered through this process. Artificially intelligent machines are now able to both support and in limited instances outperform conventional mathematics. In this talk, I will give an overview of this burgeoning field and our ongoing research by focusing on two case studies. I will present our AI guided conjecture generation framework which has been used to discover new results in multiple domains of math. I will also show how techniques such as supervised learning and symbolic regression can be utilized for this purpose in the context of string geometry.

Bio:

Daattavya’s research interests are at the intersection of mathematical physics and machine intelligence. His main focus is on developing tools for the discovery of new mathematics and analyzing their structure. Other ongoing work includes studying interesting geometries that arise in string theory and mathematical physics, often through the application of machine learning techniques. Before joining Cambridge, Daattavya graduated with an MSc. in Mathematical and Theoretical Physics from the University of Oxford where his dissertation was on Calabi-Yau Manifolds and Mirror Symmetry.

Abstract: The rapid increase of global mean temperature and unprecedented heat events require new approaches to support and monitor the climate adaptation and heat resilience of cities. Crafting effective plans necessitates accurate data and tools that adapt to the ever-changing dynamics of urban environments.

This presentation will show the recent advances in diagnosing and treating accurately, city by city, overheated urban areas (in time and space) where climate adaptation should be prioritised to promote heat resilience. The research aims to fully integrate crowdsourced urban climate observations (citizen weather stations) with satellite and remote sensing data using machine learning techniques to generate high spatio-temporal resolution observations of urban atmospheric states and dynamics. The results will support the development of an urban heat diagnosis tool with global applicability to enable insight and evidence-supported actions to promote zero-carbon and sustainable cooling at different scales. This research is part of the Future of Cooling Programme of the Oxford Martin School.



Bio: Jesus Lizana is Associate Professor in Engineering Science at the University of Oxford, with a unique experience profile in architecture and engineering. His research focuses on the cross-disciplinary challenges to support the transition towards zero carbon climate-responsive buildings.

At Oxford, Lizana is engaged in many research initiatives and has received several prestigious and extensive grants, including a Marie Curie Fellowship. He leads the research on Zero-Carbon Space Heating and Cooling at ZERO Institute and supports the interdisciplinary research in the Future of Cooling Programme of the Oxford Martin School. Alongside his academic career, Lizana also serves as a consultant on many building energy-related projects, data science, and sustainable cooling across various global locations, including the United Kingdom, India, Spain, Morocco, and Saudi Arabia.

Lizana received his PhD in low-carbon buildings at the University of Seville in Spain after completing a BSc in Architecture and an MSc in Building Engineering. Previously to his appointment at Oxford, he has lectured and conducted research at the University of Seville (Spain), the University of Edinburgh (Scotland), the Technical University of Munich (Germany), Universidade de Lisboa (Portugal), and the Spanish National Research Council (Spain).



This is a hybrid meeting. Please find the Teams link on XML webpage https://users.ox.ac.uk/~ndog0178/XML/xml_index.html.

At the core of OCLW's new programme on Writing Jewish Women's Lives, our new series of afternoon literary seminars are a chance to discuss books by and about Jewish women.



At this event we welcome Meriel Schindler to talk with Rebecca Abrams about their various books and the issues raised when writing about or as Jewish women. Topics will include family relationships, assimilation and legacy - as well as addressing practical questions about researching, writing and structuring a multiple-strand story. There will be time for questions from participants, with priority given to in-person participants.

Register here

Title: Moral Uncertainty in Autonomous Agents



Abstract:

One important goal of machine learning is to create agents that can behave ethically. Such ethical behaviour would allow autonomous agents to be safely used in a wider range of cases, e.g. fully autonomous vehicles may encounter unexpected moral dilemmas during deployment. At the same time, there is no agreement within or between societies as to what constitutes ethical behaviour. Fortunately, there has been recent work in philosophy called 'moral uncertainty', e.g. here in Oxford by William MacAskill, that aims to address this very issue. In this talk, we first look at a paper by Ecoffet and Lehman (from Open AI) to see how ideas from moral uncertainty can be used alongside with Reinforcement Learning to implement autonomous agents with multiple ethical theories. We also look at my recent AAAI paper, which looks at fanaticism, a well-known problem within moral uncertainty. In particular, should small parts of society be able to completely determine what ethical behaviour means (such that the other parts of society are disregarded), just because they are morally repulsed? Consider that only a very small part of society believes that abortion is ethically unaccaptable. Due to fanaticism, agents may also consider abortions ethically wrong, despite the majority of society thinking otherwise; is this desirable? Finally, we look at interesting unsolved problems within moral uncertainty and their relevance to machine learning.



Bio:

I am Jazon Szabo, currently a final year PhD student at the Safe & Trusted AI CDT, co-organised by King's College London and Imperial College London. My main research interests are value alignment and machine ethics; I look at the use of moral uncertainty to address problems within these fields. I am also greatly interested in the field of AI Ethics: I have been running an online AI Ethics Reading Group for 3 years and I have helped create course material for an AI Ethics course at KCL.



This is a hybrid meeting. Please find the Teams link on XML webpage: https://users.ox.ac.uk/~ndog0178/XML/xml_index.html