UC Irvine

Recent advance in artificial intelligence and machine learning, especially transformer-powered foundation models such as the Bidirectional Encoder Representations from Transformers (BERT) and Generative Pre-trained Transformers (GPT), have revolutionized how people understand and interact with information. As human filter wild world information and communicate in a set of rules and symbols that we call languages, large language models (LLMs) first emerge as our hope for intellectual agents' equivalent. Versatile as they seem, such as autocompleting code and decoding ancient languages from archaeological stones, whether LLMs can perform what we define as intelligent tasks -- such as math, planning, and reasoning -- is still debatable. This study explores the application of fine-tuning LLMs to predict user preferences, with specific focuses on the Movie industry. We call it the FiLM project (Film-interest Language Model). FiLM aims to predict the market preference for a film and identify high-propensity user segments based on early-stage movie synopses, offering a novel approach compared to traditional recommendation systems...