Adam Smith
Washington, District of Columbia, United States
Actions
Links
Area of Expertise
Topics
Aether Keys : An Intro to Backdoors in Language Models
At this point, we have all heard of at least one language model, like ChatGPT. Fundamentally, they enable users to store words as numbers & compare those numbers in a novel way to form expert level responses to a user's query about very complex data sets. While a welcome addition to most people's workflow, it is quickly becoming apparent how very little the average user understands the risks that accompany its use.
Only in the past 2 years has research into model focused attacks begun to emerge. In that time, most research assumes a very deep understanding on the reader's part, while little has been done to describe the state of the field in general.
So, how are you supposed to know how to evaluate a model's risk, understand model attack paths, or detect a trojanized model? This talk aims to answer all of those questions, by giving a description of the state of language model security, & provide a model of exploitation tactics, techniques, & tools.
Adam Smith
Washington, District of Columbia, United States
Links
Actions
Please note that Sessionize is not responsible for the accuracy or validity of the data provided by speakers. If you suspect this profile to be fake or spam, please let us know.
Jump to top