Profs and Pints: Baltimore presents “The Weaponization of Generative AI,” on the manipulation of artificial intelligence to harm others and how it can be countered, with James Foulds, associate professor, and Philip Feldman, adjunct research assistant professor, from the Department of Information Systems at the University of Maryland, Baltimore County.

On May 14th Grok, a generative artificial intelligence chatbot developed by Elon Musk’s xAI Corporation, began talking about “White Genocide” in South Africa. Officials at xAI blamed a “rogue employee,” but on July 8th the AI chatbot was at it again, calling itself “MechaHitler” on the X platform and posting there antisemitic content and praise of Adolf Hitler—comments echoing opinions previously voiced by Musk.

These incidents weren’t just offensive but also alarming, because they demonstrated how generative AI technologies can be intentionally manipulated to produce misleading, ideologically motivated, or otherwise harmful content—or, in a word, weaponized.

Join Dr. James Foulds, an expert on the responsible use of artificial intelligence, and Dr. Philip Feldman, an expert on AI’s misuse, for a look at the weaponization of AI and how it can be stopped.

Dr. Foulds will begin with a primer on the technology underlying AI and how features intended to prevent AI behaviors like those in the recent Grok incidents were instead harnessed to cause them. He’ll help you gain a basic understanding of the large language model (LLM) technology that underlies generative AI chatbots such as Grok and ChatGPT and the methods used to align AI with human values. He’ll discuss how AI alignment was misused to steer Grok to produce propaganda, and how this illustrates how AI chatbots can be weaponized to try to influence and control society’s institutions.

From there, Dr. Feldman will talk about how generative AI has created new opportunities for information warfare. He’ll describe how chatbots have already manipulated people to commit violent acts accidentally, and he’ll discuss how deliberate “weapons-grade” manipulation could inflict significant damage on individuals, organizations, and even nations. We’ll examine the possibilities of large-scale AI-driven psychological manipulation (PSYOPS).

The speakers will conclude by discussing strategies and interventions to counter the emerging threat of AI weaponization and steps you can take to protect yourself. (Advance tickets: $13.50 plus sales tax and processing fees. Doors: $17, or $15 with a student ID. Listed time is for doors. The talk starts 30 minutes later.)

Image by Canva.