Generative AI-in-the-loop: Integrating LLMs and GPTs into the Next Generation Networks

This paper investigates the integration of large language models (LLMs) like GPT into next-generation mobile networks, particularly in the context of 6G, under the concept of "generative AI-in-the-loop." LLMs are proposed to complement traditional ML algorithms, addressing their limitations in handling complex scenarios. The study analyzes LLM capabilities, emphasizing their understanding, reasoning, and ability to generate synthetic data. A case study on enhancing intrusion detection with synthesized data showcases improved performance. The paper discusses the benefits of combining LLMs and ML models, such as better network management, automation, and adaptability, while addressing challenges like data scarcity and privacy concerns. The integration can take place in centralized, distributed, or hybrid forms, with LLMs assisting in various stages of the ML model lifecycle, from data processing to model evaluation. The use of LLM-generated synthetic data, though promising, requires careful quality assessment. The research aims to bridge the gap between LLMs and ML-driven networks and paves the way for more advanced, AI-native systems in mobile communications.