The 6th International Conference on Modern Management based on Big Data (MMBD 2025)

Abstract: The power of Large Language Models (LLM) has opened new possibilities for businesses, ranging from chat bots for customer and employee support to instruments for business intelligence. These applications promise increased profit, better customer service, higher-quality prognosis, etc. As such the number of applications of LLM is growing seemingly without limit. However, LLM present a number of weaknesses and risks of which the vast majority of users are unaware, including bias, "hallucination" (i.e., inventing fake "facts"), and even cybersecurity risks. Fortunately, there exist technologies and means of mitigating these risks.
In this talk I will show how LLM work (for the lay person) and why these risks occur. I will also discuss uses of Retrieval-Augmented Generation (RAG) and other technologies for addressing problems. As well, you will learn when it is acceptable to use a pre-trained model and when you really need to train your own.

Biography: Dr. Jennifer Q. Trelewicz is CEO and Founder of TGPO Consult, Ltd., and has clients in major corporations in the telecommunications and transport sectors. She has been working in the field of AI and machine learning since the 1980s, information security since the 1990s. She has more than 30 years of experience in technical and top-management positions in leading companies in their field, including IBM, Google, Microsoft, Mail.Ru, Deutsche Bank Tech Centre, Motorola, and others. She has publications in refereed journals and conferences internationally, 28 issued technology patents in various countries. She is a senior member of IEEE and a member of the IEEE committee for publications.