Mediating human-machine collaboration for creativity-based learning is investigated, emphasizing the use of search-based procedural content generation (PCG) technologies, like Map-Elites, alongside generative adversarial networks (GANs). Current collaboration systems lack personalization and adaptability due to limited autonomy and data. A meta-learning technique is proposed to address these challenges, enabling effective autonomy with limited data. The research aims to develop a co-dependent system based on GANs and Map-Elites, automate the collaboration framework using meta-learning, and design an evaluation scheme to maximize the framework's performance. The project aims to create an innovative creativity-based learning framework, potentially transforming education and knowledge acquisition.

The integration of AI technology and Entertainment Science has revolutionized gaming experiences, particularly in the wake of the COVID-19 pandemic. This collaboration has become increasingly vital across various industries. The project's goal is to establish empathetic entertainment as a robust research field that can provide insights into human interactions and behaviors.

The integration of AI technology in gaming has brought about significant advancements in research areas such as game-solving and general game-playing (GGP). However, there is a need for more empirical verification of the concept of "motion-in-mind" due to its subjective nature. While AI can assist in identifying enjoyable aspects of games, it still lacks the ability to comprehensively assess levels of distress and break down complex events with imperfect information. To address these limitations, this project aims to combine AI research and entertainment science to establish a new field of study called "comfort in mind" through games.

The World Health Organization reported that the number of people affected by depression has increased in recent years. However, one of the contributing factors that are least investigated is global warming, which has raised the hospital admission rate of depression patients. This study explores the impact of global warming concerning age group and residency on the upswing rate of depression patient hospital admission. The study's objectives are to formulate the relationship of global warming, place of residence, and age to the increasing rate of depression patients, determine the ambient temperature that impacts depression patients adversely, and propose and evaluate a hybrid of DRNN-HMM approach to predict the rising rate of depression patients. The expected implication of the study is a state-of-the-art machine learning approach that can formulate and alert the impact of global warming on depression patients in Malaysia.

# RLHF is just barely RL

Reinforcement Learning from Human Feedback (RLHF) is the third (and last) major stage of training an LLM, after pretraining and supervised finetuning (SFT). My rant on RLHF is that it is just barely RL, in a way that I think is not too widely… pic.twitter.com/sjRZvqc5KC

— Andrej Karpathy (@karpathy) August 7, 2024