Cohere For AI Invited Talk

Image credit: Cohere


In today’s world, the rapid advancement and widespread use of large language models (LLMs) have brought about both opportunities and challenges. While these models have tremendous potential, they also pose risks, particularly in generating harmful and misleading content. Our research introduces an innovative “Fighting Fire with Fire” (F3) strategy to address this issue by leveraging the capabilities of modern LLMs. Our approach utilizes GPT-3.5-turbo to generate both authentic and deceptive content through advanced paraphrase and perturbation techniques. Furthermore, we apply zero-shot in-context semantic reasoning to differentiate genuine from deceptive posts and news articles. Our extensive experiments demonstrate that GPT-3.5-turbo achieves a superior accuracy of 68-72% in detecting disinformation, outperforming previous models. This research underscores the potential of using advanced AI to combat the very problems it may create. In another study, we address the spread of COVID-19 misinformation in low-resource regions, focusing on the Caribbean. Given the lack of abundant fact-checking resources in these areas, we transferred knowledge from fact-checked claims in the US to detect misinformation in English, Spanish, and Haitian French. Our findings highlight the challenges and limitations of current fake news detection methods in low-resource settings and emphasize the need for further research in multilingual detection.

Apr 11, 2024 9:00 AM — Apr 13, 2024 9:20 AM
Online Presentation
University Park, Pensylvania 16802-2440
Click on the Slides button above to view the built-in slides feature.

Slides can be added in a few ways:

  • Create slides using Hugo Blox Builder’s Slides feature and link using slides parameter in the front matter of the talk file
  • Upload an existing slide deck to static/ and link using url_slides parameter in the front matter of the talk file
  • Embed your slides (e.g. Google Slides) or presentation video on this page using shortcodes.

Further event details, including page elements such as image galleries, can be added to the body of this page.

Jason Lucas
Jason Lucas
Ph.D. Student in Informatics

My research interests include low-resource multilingual NLP, linguistics, adversarial machine learning and mis/disinformation generation/detection. My Ph.D. thesis is in the area of applying artificial intelligence for cybersecurity and social good, with a focus on low-resource multilingual natural language processing. More specifically, I develop NLP techniques to promote cybersecurity, combat mis/disinformation, and enable AI accessibility for non-English languages and underserved populations. This involves creating novel models and techniques for tasks like multilingual and crosslingual text classification, machine translation, text generation, and adversarial attacks in limited training data settings. My goal is to democratize state-of-the-art AI capabilities by extending them beyond high-resource languages like English into the long tail of lower-resourced languages worldwide. By innovating robust learning approaches from scarce linguistic data, this research aims to open promising directions where AI can have dual benefits strengthening security, integrity and social welfare across diverse global locales.