Events

AI assistants (large language models; LLMs) are an ever-increasing presence in our classrooms, labs, and research groups. They increase our productivity, but also bring challenges to the way we learn and do research. In this seminar we will discuss ButterflAI, a learning experience designed with one technical goal: Creating a statistical model that can generate synthetic butterfly diagrams. Besides our technical goal, our program had multiple educational goals: 1. Leverage AI assistants to address the needs of a very diverse population in terms of technical proficiency and specialized knowledge (highschool to professional scientist). 2. Transform AI assistants from task oriented to education centered tools. 3. Evaluate the benefits of AI assisted research collectives, where the intersection of human curiosity and AI randomness enlarge the discovery space and produce better research outcomes.

We will start by discussing the conceptual framework underpinning our use of AI assistants (striving to equal human-AI partnership). Then we will discuss the structure of the program and the rationale behind some of our program design choices, and we will finish with a summary of the technical achievements which are: 1. The implementation of a competitive solar cycle generator using classical statistical modeling. 2. The augmentation of the classical model using AI-assisted statistical approximations (reverse diffusion).

Speakers and their affiliations:

1. Andrés Munoz-Jaramillo, Southwest Research Institute
2. Phoebe Mahlin, Space Sciences Laboratory, University of California, Berkeley
3. Claudia Holland, Space Sciences Laboratory, University of California, Berkeley
4. Swaksh Udupa, COFFIES
5. Myra Khieu, University of California, Berkeley
6. Jessica Hamilton, Georgia State University
7. Oana Vesa, Stanford University
8. Allison Constantini, COFFIES