Despite numerous advantages, LLMs have trust, transparency, accountability, and reliability issues due to development with "black-box" approaches, which make it difficult to understand how LLMs create specific outputs.
Trustworthy LLMs: Principles and Challenges presents the fundamental concepts of trustworthy LLMs, then proceeds to address the foremost challenges researchers and developers face in developing reliable and trustworthy LLMs. The book begins by presenting the main research branches of artificial intelligence along with the principles of LLMs, from pre-training to fine tuning, and, ultimately, trustworthy LLMs. Readers will learn about the chief technical principles of LLMs, including attention mechanism, transformers, and transfer learning. The methodologies used for development of ChatGPT have been explained as a case study for comprehensive understanding of the concepts involved in LLMs. Readers will also learn about the integration of XAI with LLM, and other key frontiers in trustworthy LLM development, including the synergy between deep learning and LLMs, as well as case studies on GPT-4 and OPT-1.3B. The book concludes with chapters on key challenges and future research approaches for developing trustworthy LLMs.
- Provides a well-organized framework to understand how Large Language Models work, and how they are evolving
- Offers a structured mapping of LLM methodologies, including transformers, attention mechanism, and transfer learning
- Presents the latest technical frameworks combined with real-world examples and case studies such as ChatGPT, TrustGPT, GPT-4, and OPT-1.3B
- Provides a comprehensive reference on trustworthy LLMs, along with the challenges and future research frontiers for their development
