Generative AI including Large Language Models for Social Good
This talk will explore an application of generative AI, specifically Large Language Models (LLMs), in the domains of political conflicts. Dr. Khan and his research team conducted global monitoring of conflicts and political violence by analyzing vast amounts of specialized text. Collaborating with a multidisciplinary team, the team developed ConfliBERT, a domain-specific pre-trained language model focused on conflict and political violence in English, Arabic and Spanish. ConfliBERT is publicly available on Github and Hugging Face. Since its recent release, the models have been downloaded over 14,000 times from Hugging Face, indicating significant interest. During this talk, Dr. Khan will demonstrate how their model outperforms standard LLMs in various tasks such as classification and question answering.
Speaker biography
Dr. Khan is an international leader in Big Data Analytics (BDA), a key aspect of data science. He has been instrumental in educating the students in BDA at UTD, for several years. He was also successful in writing multiple proposals to NSF and getting funding for experimental research and introduced his BDA students to research projects. His BDA class has often had enrollments of around 130 and is extremely popular. He has also developed a BDA framework and collaborates with faculty at EPPS (e.g., Profs. Jennifer Holmes and Patrick Brandt) and shares his framework with them for their applications in political science. Together with EPPS professors as well as through other collaborations (e.g., UIUC, U of MN) he has brought in millions of dollars in federal funding in this area. He is known worldwide as the Data Science person at UTD by eminent researchers.