SUMMER RESEARCH FELLOWSHIP 2022

The 2022 Summer Research Fellowship was a 12 week program designed to introduce fellows to the field of existential risk, build research experience and career capital, and demonstrate the potential for a career in the field. The 10 fellows of the initial cohort designed and conducted their own research projects under the mentorship of a faculty member or senior researcher. All 10 fellows focused on AI or disinformation. Over the course of the summer, fellows attended workshops on:

Workshop on How to Plan a Research Project
Theory of Change Workshop w/ Michael Aird, Rethink Priorities
Modeling Existential Risk w/ Professor James Evans
Nuclear and Climate Risks w/ Professor Daniel Holz
Workshop on Research Productivity
Q&A w/ Peter Wildeford, Co-CEO of Rethink Priorities
Knowledge in AI Models w/ Professor Victor Veitch
Workshop on Preventing Burnout
Conflicts between Science and Public Perception w/ Professor Robert Rosner
Interpreting AI Risk in Relation to Biological Threats w/ UChicago Office of Research Safety

2022 FELLOWS AND THEIR PROJECTS

SANDRO DEL RIVO

UChicago Computer Science & Philosophy ‘23
Mentor: Victor Veitch (UChicago)

Project: Will AGI Learn Alien Concepts?
An exploration of how advanced AI models might form representations of objects and concepts

SOREN DUNN

UChicago Chemistry, Statistics, & Data Science ‘23
Mentor: Victor Veitch (UChicago)

Project: Improving Evaluation of Factual Accuracy in Language Models
Investigates how to ensure AI models output true statements and not false but true-seeming or made-up responses

GRANT HARRIS

UChicago Economics & Computational and Applied Math ‘25
Mentor: Ali Merali, Economics PhD Candidate, Yale University

Project: Cooperative AI: A Primer
A literature review of game theoretic implications for multi-agent AI models and human-AI interaction

SUDHAMSHU HOSAMANE

UChicago MA Computational Social Science ‘23
Mentor: Munmun De Chaudhary, Georgia Institute of Technology

Project: Effective Strategies to Tackle Misinformation on Messaging Apps
Proposal for a study of how misinformation travels on encrypted messaging apps and intervention effectiveness

HENRY JOSEPHSON

UChicago Philosophy & Data Science ‘25
Mentor: Cullen O’Keefe (OpenAI Research Scientist)

Project: A California Effect For Artificial Intelligence
Exploration of the feasibility for regulatory diffusion under various regulatory targets in California

THEO KO

UChicago Computer Science & Computational and Applied Math ‘24
Mentor: Claire Donnat (UChicago)

Project: Accessing the Fairness Level of the Training Set of a Blackbox Graph Neural Network Model through Node Masking
Study of whether different nodes in an ML model have predictably greater influences on outputs

NICOLE MATEO

UChicago Psychology & Critical Race and Ethnic Studies, Data Science Minor ‘24
Mentor: Anna Reosti (American Bar Foundation)

Project: Survey of POC and Impressions of AI Risk
Investigating how people of color consider the importance of existential AI risk given early AI applications have demonstrated harms to minorities

JACK WIPPELL

UChicago MA in International Relations ‘22
Mentor: Robert Trager (UCLA), Duncan Snidal (Oxford)

Project: When Deviation is Not an Option: Lessons from Reykjavik on the Possibility of Zero-Fail Treaties
Research into whether treaties can be designed and enforced when deviation risks catastrophe

ANDREW YANG

UChicago Computer Science & Linguistics ‘23
Mentor: Allyson Ettinger (UChicago)

Project: Knowledge Graphs for Language Model Evaluation
Suggestion that knowledge graphs be used to understand how concepts are encoded in language models

FREDDIE YANG

MA with Certificate in Research Methods (UChicago Harris School of Public Policy) ‘23
Mentor: Allyson Ettinger (UChicago)

Project: A Random Thought On Multi-task Learning And AI Existential Risks
Investigation into whether training on one task can positively influence performance on another