Generative AI & LLMs

This session provides a comprehensive overview of artificial intelligence (AI), focusing on the evolution, concepts, and applications of Generative AI. The session explores topics such as the history of AI, different types of machine learning, neural networks, deep learning, large language models, and AI agents. It examines various generative models, including Generative Adversarial Networks (GANs) and Variational Autoencoders (VAEs). The session emphasizes prompt engineering and transfer learning techniques for optimizing the performance of these models.

Slide deck posted on the iCollege class site.

Suggested Reading

• Textbook Hands-On Large Language Models by Jay Alammar and Maarten Grootendorst
  • Chapter 1. An Introduction to Large Language Models
  • Chapter 6. Prompt Engineering
• Textbook Fundamentals of Machine Learning for Predictive Data Analytics by John D. Kelleher, Brian Mac Namee and Aoife D’Arcy
  • Chapter 8. Deep Learning

Highlights

This session emphasizes several key insights about Generative AI:

• Generative AI is a rapidly evolving field with the potential to revolutionize numerous industries. Generative AI models, trained on vast datasets, can produce novel and realistic content, including text, images, videos, and audio. Applications range from content creation and drug discovery to personalized recommendations and even generating creative text like poems or code.

• Large Language Models (LLMs) are foundational to Generative AI. LLMs are trained on massive text datasets using self-supervised learning techniques, enabling them to understand and generate human-like text. Architectures like transformers, incorporating attention mechanisms, facilitate capturing long-range dependencies in text.

• Fine-tuning and alignment are crucial steps in developing effective and responsible LLMs. Fine-tuning tailors pre-trained models to specific tasks, while alignment techniques, often involving human feedback, ensure model outputs adhere to human values and ethical standards.

• Prompt engineering plays a vital role in harnessing the capabilities of LLMs. The quality of prompts directly impacts the relevance and accuracy of model outputs. Effective prompts provide clear instructions, context, and desired output format, guiding the model towards producing desired results.

• Transfer learning is a powerful technique that leverages pre-trained models to expedite and enhance the development of new models. This approach applies knowledge gained from one task to another related task, reducing training time and improving performance. Transfer learning finds applications in both computer vision, such as image classification and object detection, and natural language processing, such as sentiment analysis and language translation.

• The development of AI agents, powered by LLMs and augmented with external knowledge, holds significant potential for creating intelligent systems capable of autonomous task execution and human-like interaction.. These agents can access external databases, adapt to dynamic environments, and continuously learn, making them versatile for applications ranging from personal assistants to components of complex autonomous systems.