Decentralized autonomous agents with LLM
Dawn Song, Professor
Electrical Engineering and Computer Science
Closed. This professor is continuing with Fall 2023 apprentices on this project; no new apprentices needed for Spring 2024.
As we tread the path of rapid evolution in the domain of Language Learning Models (LLM), this project seeks to address the major challenges of integrating personal privacy into this rapidly expanding field. The critical question we aim to answer is: how can LLMs safely and effectively utilize private user data to enhance their functionality, and what ramifications does this have on their behavior and interaction within a decentralized system?
Role: We intend to build an advanced decentralized autonomous agent that functions based on user data, maintaining the users' privacy while fostering a highly personalized interaction. The research will delve into the development of sophisticated algorithms that will enable the LLM to use private data in a secure manner. For instance, we aim to develop an LLM agent capable of recommending personalized content based on a user's browsing history, all while safeguarding the individual's privacy.
Moreover, a distinctive aspect of this project will focus on understanding the evolution of LLMs in a decentralized setting when private data and context are incorporated. With each LLM agent having its unique set of data, they will behave differently, potentially leading to the formation of a decentralized autonomous community. This, in turn, raises intriguing questions about the communication behavior between LLM agents functioning within asymmetric knowledge contexts. The research will analyze these interactions. The resultant exploration could provide valuable insights into LLM-based agent design and its practical applications.
Qualifications: A foundational understanding of programming in JavaScript, Python, or Rust. A basic grasp of natural language processing, specifically regarding language models and their functioning. An interest in security and privacy-related subjects is highly encouraged, and any related experiences would be advantageous.
Hours: 12 or more hours
Engineering, Design & Technologies