ChatGPT vs. Edward SAID: Generative AI in Underresourced Domains

Ever tried working with ChatGPT but it didn't seem to know anything about Abbasid panegyrists or Ottoman legal records? Or did it just not want to talk to you at all because your inquiry was deemed too politically sensitive? Join us for a 2-hour workshop exploring the impact of Large Language Models (LLMs) on underresourced languages and research domains! We will discuss how the training data and architecture of LLMs can amplify biases and misinformation and explore how technical concepts like probabilistic information retrieval and transfer
learning affect our scholarship in such fields. We’ll learn to ask the right questions and examine different methods of prompt engineering to mitigate misinformation and misrepresentation.

And if you’re tired of only getting generic answers - or none at all, - throw in a "RAG" (Retrieval Augmented Generation) and watch it spin your queries into tailored answers. All problems solved? Nope. But let’s join the discussion on how we can responsibly engage with this technology, contribute to its ethical development, and replace PhD advisors. Not.

Note: No prior knowledge of any kind required. Maximally 15 participants.
Registration required.

Georgetown Arabic Department Graduate Association | View More Events

Contact the organizers