Getting started with Generative AI – 2 hours
- Understanding Generative AI
- Types of Generative Models – autoregressive models, variational autoencoders, and generative adversarial networks (GANs)
- Categorizing generative models based on learning algorithms: likelihood-based vs. likelihood-free
- Motivation for generative modeling compared to discriminative models
- Characteristics of generative models: density estimation, data simulation, representation learning
- Lab(s): Getting started with Generative AI Models
Large Language Models (LLMs) – 4 hours
- Introduction to LLMs
- Use cases and tasks of LLMs
- Architecture of LLMs
- Generative Models for Text: Introduce generative models for text generation in NLP, including approaches like language modeling with autoregressive models, variational autoencoders (VAEs), and transformers.
- Evolution of text generation techniques
- Understanding role of Vector Databases
- Prompting and Prompt engineering
- Lab(s): Prompt Engineering
Evaluating LLMs – 3 hours
- Evaluating LLMs Significance and impact of Evaluation on natural language understanding and generation tasks
- Various Evaluation Metrics used to assess the quality and performance of LLMs
- Perplexity,
- BLEU,
- ROUGE,
- METEOR, and others commonly used in machine translation, summarization, and text generation tasks
- Human Evaluation in assessing LLMs
- Intrinsic & Extrinsic Evaluation
- Dataset Quality and Bias
- Interpretability and Explainability
- Robustness and Generalization
- Evaluation in Low-Resource and Multilingual Settings
- Fairness and Bias Evaluation
Fine Tuning Basics – 2 hours
- Background and concept
- Curse of dimensionality
- Graphical models (Bayesian networks)
- Comparison of generative and discriminative models
- Lab(s): Fine Tuning for specific tasks
Scaling Human Feedback – 2 hours
- Challenges and considerations in scaling human feedback
- Strategies for collecting and incorporating large-scale feedback
