Anna Kircher - Analyzing Consistency in LLM Outputs Leveraging Colourful Queries

Analyzing Consistency in LLM Outputs Leveraging Colourful Queries by Anna Kircher
Visit rstats.ai for information on upcoming conferences.
Abstract: Approaching the GenAI’s black box with the power of colours. My illuminating journey through the mysteries of colour symbolism and its interpretations using ChatGPT and R. Through specific queries about metaphors, saying and meanings of colours, responses are generated. Analysis of said responses is conducted to further interpret and summarize colour perception and interpretation, shedding a little light on the intricate nuances of colour symbolism. By introducing randomness through temperature variation in underlying language models, the creative potential and consistency in responses is explored and shifts in the interpretation of colour symbolism uncovered.Overall leading to unraveling the intersection between language, colours, perception and AI.
Bio: Anna is a Senior Consultant at EY’s AI & Data Advisory Services practice, specializing in risk analytics and ESG and climate related data modeling, integration and reporting. With over 6 years of experience in the financial service industry in developed and emerging and frontier markets.
Throughout her career, Anna has thrived on diverse projects, engaging in qualitative analysis, crafting risk assessment frameworks, supporting macroeconomic and climate scenario modeling and analysis, as well as enabling faster and data-driven decision-making through statistical modeling and machine learning. She holds a Master’s degree in Mathematics and is passionate about colours, R and crocheting.
Presented at the 2024 New York R Conference (May 17, 2024)
Hosted by Lander Analytics (landeranalytics.com)