I am studying public health in Barcelona, but I struggle with Spanish and, clearly, with the material. In between understanding the nuances of the language and the material, I have been staring at this assignment for over a week. Would you please help me understand whom/what are the prompts (in step 2) meant for–are they directed towards the application?
This practicum aims to provide a solid understanding of the importance of prompts in generating text with LLM, and how careful articulation of these can be a catalyst for extracting value from AI and NLP in the health domain. The aim is to equip participants with the necessary skills to design, evaluate and optimize prompts that enable LLMs to be useful and effective tools in the promotion of informed health, based on data supported by the CRISP-DM methodology.
(1) Business Knowledge - Thematic Exploration in Digital Health: You will investigate a relevant topic in digital health related to mental health - suicide or addictions, such as electronic health records, telemedicine, health data privacy or mobile health applications, to give some examples.
(2) Understanding and Preparing the Data - Creating Focused Prompts: You will design five prompts that address different aspects of the selected topic, focusing on how each aspect can be explored or improved through AI. Explain best practices for creating prompts in healthcare.
(3) Modeling - Implementation with LLM: You will use different Large Language Models, such as BARD and ChatGPT, to generate responses to the prompts created, exploring the variety and quality of responses generated by each model.
(4) Evaluation - Responses: You will evaluate and compare the relevance, accuracy, and consistency of the responses from each model, identifying strengths and weaknesses in text generation in relation to the digital health topic explored.
(5) Deployment - Reflection and Prompts Improvement: Will reflect on the findings, discussing how prompts could be improved to deploy in a healthcare setting, elicit more accurate and relevant responses from LLMs, and how these findings can inform future practices in digital health-oriented text generation.