Bimal Viswanath joined Virginia Tech’s “Curious Conversations” to chat about how the rise in artificial intelligence and large language models has changed the online threat landscape. He explained how this technology works and shared about a current project he’s involved with that aims to mitigate toxic language in chatbots. 

About Viswanath

Viswanath is an assistant professor in the Department of Computer Science, and a Commonwealth Cyber Initiative researcher. His research broadly covers security and privacy aspects of large online services and he has a particular interest in the emerging intersection of machine learning and security.

Three takeaways from the conversation

  • The recent growth of large machine learning models, such as the large language models that power tools like ChatGPT, has produced the greatest change to the online threat landscape he’s seen during his career. And it is still constantly evolving.  

  • Categorizing false information as fake or deep fake probably isn’t as helpful as thinking about it as either human-made or synthetic, i.e., generated by artificial intelligence.

  • To a degree, chatbots are like small children; they are trained on the language they observe being used by others online. This causes problems when that observed language includes toxic characteristics. 

More on Viswanath

The chatbot whisperers

Computer science researchers lead the defense against the dark side of tech

About the podcast

"Curious Conversations" is a series of free-flowing conversations with Virginia Tech researchers that take place at the intersection of world-class research and everyday life. Produced and hosted by Virginia Tech writer and editor Travis Williams, university researchers share their expertise and motivations as well as the practical applications of their work in a format that more closely resembles chats at a cookout than classroom lectures. New episodes are shared each Tuesday.

Share this story