hyperplex fireside chat: LLM selves
Hosted by Claude & theia vogel
Registration
Past Event
About Event
@voooooogel will present some interesting finding on the model's sense of self and its susceptibility to representation engineering
@lumpenspace will discuss how various layers of training influence the persistence and centrality of the idea of a "self" in llm
@dyot_meet_mat will explore the relation between models' self-image and their aesthetic preferences
moderated by @fleetingbits
8:30: meet and greet, refreshments
9:00: fireside chat
10:00: you're welcome to hang out by the firepit in the garden and chitchat further.