Artificial Intelligence II

Module 1<br> This module builds on the foundations established in Artificial Intelligence I and introduces advanced methods in machine learning. The module covers stochastic and probabilistic ML techniques, deep reinforcement learning including deep Q-learning and policy-based methods, and associated architectures. Advanced neural network architectures such as LSTM, RNNs, transformers, and diffusion models are studied. The module also introduces the theoretical principles of modern generative AI, including large language models and multimodal models. Students learn to critically evaluate AI models using appropriate metrics and assessment methodologies, and to select appropriate models for different application domains. <br> <br> Module 2<br> This module focuses on the practical challenges of training, fine-tuning, and deploying advanced AI models at scale. Students study optimisation techniques such as quantisation and efficient fine-tuning techniques such as LoRA, and prompt tuning. The module covers the adaptation of large-scale pre-trained models to domain-specific tasks, including instruction tuning, preference optimisation (e.g., RLHF, DPO), and retrieval-augmented generation (RAG). Students explore applied AI systems that integrate model capabilities with external tools, memory mechanisms, and retrieval pipelines. Students gain experience implementing these methods using frameworks such as PyTorch. The module concludes with a critical examination of the ethical and societal implications of AI, covering topics such as bias and fairness, transparency and explainability, intellectual property, environmental impact, AI alignment, and responsible AI development. These discussions are grounded in the systems and techniques developed throughout the module.