ÍÃ×ÓÏÈÉú

Bulletin Description:

This course introduces students to AI tools that allow the creation of new data such as text (Natural Language Generation), images, or videos. Students use Large-Scale Language Models (LLMs) to generate text, evaluate its quality, and integrate the generator in sample applications. Students also create images via deep neural networks, although no prior familiarity with deep learning is required. Students will complete a term project.

Prerequisite:

CSE 374

Learning Outcomes:

1. Describe how generative AI relates to other areas of machine learning and can be used to perform specific tasks.
2. Implement Python code to generate data. (a) Apply modern programming environments (e.g., Jupyter Notebooks, APIs from modern libraries) to use Generative AI tool and generate text, images, and/or numerical data. (b) Describe key concepts in data generation, including oversampling, fairness and bias, conditioning, alignment, and hallucination.
3. Evaluate the quality of generated AI solutions using quantitative and qualitative metrics. (a) Implement and apply quantitative algorithms to evaluate the quality of generated data, such as overlap, semantic similarity, or n-gram matches. (b) Apply qualitative assessment processes (e.g., for fluency, faithfulness, coverage) and determine the inter-rater agreement.
4. Optimize existing AI Solutions to deliver expected results. (a) Use prompt engineering and fine-tuning to identify the best generator’s hyper-parameters to generate an AI output that performs best on an application scenario (e.g., question-and-answering system). (b) Extend a knowledge model with external data.
5. (Graduate students only) Adapt methods and metrics from recent research articles (beyond what is covered in class) to develop robust, fine-tuned, generative AI models.
6. (Graduate students only) Demonstrate an ability to do independent research in identifying and applying state-of-the-art metrics to assess and evaluate generative AI models.