Details
State of the Art in Knowledge Editing
Alex Loftus
Work across many subfields of machine learning has become increasingly reliant on billion parameter-scale models. If we can find methods to update these models without spending the time and computation necessary to run full fine-tuning sessions, we unlock customization previously only possible with industry-scale compute power. A parallel and related question is in interpretability: Where does knowledge live in these giant models? How is it stored, and can we find interpretable directions in their weight-space?
Join us for an exploration of the research surrounding these questions. We'll explore work coming out of Jacob Andreas' lab at MIT, Christopher Manning at Stanford, Jacob Steinhardt's lab at Berkeley, and Ludwig Schmidt's group at the University of Washington. We'll also look at some open-source work being done by EleutherAI as well as knowledge-editing work in natural language processing done by the Allen Institute for AI's Israeli team.
Alex Loftus is a data scientist working at a company using machine learning to speed up the drug discovery process. He holds an undergraduate degree in neuroscience with minors in chemistry and philosophy at Western Washington University, and a master's in biomedical engineering with a data science concentration from Johns Hopkins University.
Please make sure to read the instructions for joining the event below.
Agenda:
12:00 - 12:05 pm -- Arrival
12:05 - 1:25 pm -- Presentation and discussion
Time permitting -- Additional Q&A
Links to notes/slides and videos of prior meetups are available on the SDML GitHub repo https://github.com/SanDiegoMachineLea...
Location:
Please Note: There are two steps required to join this online meetup:
You must go to our Slack community and ask for the password for the meeting. Link to join is below.
You must have a Zoom login in order to join the event. A free Zoom account will work. If you get an error message joining the Zoom, please login to your account on the Zoom website then try again.
Community:
Join our slack channel for questions and discussion about what's new in ML:
https://join.slack.com/t/sdmachinelea...