Stephanie Rosenthal and Reed Simmons, professors at Carnegie Mellon University, have developed an innovative course called "Autonomous Agents: An Advanced Course on AI Integration and Deployment." This course aims to bridge the gap between AI integration and robotics application, offering students a project-focused approach to integrate AI techniques into a complete system. The course emphasizes well-structured software architecture, thorough testing, and validation - a crucial aspect of long-term deployment. Students work on an intriguing project involving autonomous greenhouses where AI agents control actuators based on sensor input to grow plants. The course covers topics like computer vision, explainability, and scheduling, challenging students to optimize their agents' performance. The ultimate goal is to create robust and reliable agents capable of autonomous plant growth. This course exemplifies the exciting possibilities of AI in real-world applications and ethical discussions surrounding autonomous systems' deployment.

Predicting outcomes, especially in finance, where investments are at stake, is a universal desire. The challenge lies in optimizing these predictions to maximize returns on investments, a critical concern for businesses and pension funds reliant on contributions. Human predictors in finance may carry their biases, creating trust issues between them and investors. To address this, a paper introduces the Random Neural Network, inspired by brain structures, for predicting time series data like prices. Unlike traditional methods, this model relies solely on historical data, offering accurate predictions for diverse investment portfolios, including property, stock, and cryptocurrency markets.

In the realm of customer service, the widespread adoption of virtual agents (bots) has often left customers dissatisfied, prompting a need for a more comprehensive approach. This approach involves the collaboration of virtual agents and human operators, combining various AI technologies such as natural language processing, multi-agent systems, machine learning, reinforcement learning, and behavioral cloning. Additionally, it calls for the integration of insights from diverse disciplines, including psychology, business, sociology, economics, informatics, and more. This paper introduces the concept of holistic customer service and explores the necessary AI technologies and interdisciplinary aspects for its success. Bridging the gap between these communities with different methodologies and evaluation criteria presents a challenge, but the potential benefits include cost-effective and improved customer service, along with fostering collaboration among diverse fields to address complex challenges.

In the context of fuzzy hypothesis testing, non-asymptotic fuzzy estimators offer triangular-shaped fuzzy numbers, providing more continuous and compact support compared to traditional estimators. This approach excels in situations requiring subtle comparisons between closely related statistical quantities. When applied to hypothesis testing, non-asymptotic fuzzy estimators outperform previous methods, granting the flexibility of partial rejection or non-rejection of the null hypothesis, especially in critical scenarios.

A network of Bayesian agents has been created to collectively model teammates' mental states based on observed communication. This innovative approach enhances the collective intelligence of human-AI teams by generating interventions that surpass individual human capabilities and introduces a real-time measure of theory of mind ability, shedding light on human cognition. The study reveals that humans often struggle to fully integrate information from teammates, especially under high communication loads, and exhibit cognitive biases that lead them to undervalue potentially useful but ambiguous information. This theory of mind ability measure predicts both individual and team performance, significantly improving our understanding of team dynamics and decision-making processes.

In the pursuit of advancing Artificial Creative Intelligence (ACI) and overcoming the imitation barrier, a hierarchical architecture known as the Insight-Knowledge Object (IKO) model is introduced as a framework. This model represents a significant step in replicating human intelligence within computational machinery.

This presents a novel approach to enhancing decision-making architectures for FIFA player agents. Leveraging Convolutional Neural Networks (CNNs) and evolutionary techniques, this research focuses on optimizing the decision-making processes in FIFA gameplay scenarios. By applying evolutionary algorithms to refine the CNN-based models, this study seeks to elevate the strategic intelligence of player agents, thereby enriching the gaming experience and pushing the boundaries of AI-driven gameplay in FIFA.

The primary objective of the DALI system is to enhance conventional traffic control systems by imbuing them with autonomy and intelligence. This is achieved by integrating a software agent into every pre-existing intersection controller, effectively transforming it into the controller's central decision-making unit. These agents are tasked with the analysis of traffic data, facilitating direct communication between one another, and working collaboratively to implement an optimized timing strategy aimed at enhancing the overall flow of traffic.

The challenge of estimating the Estimated Time of Arrival (ETA) in Intelligent Transportation Systems is a fundamental problem, involving predicting travel time along a specified route. This process relies on embedding vectors to represent route segments, but often encounters data sparsity issues, particularly for less-traveled routes. To tackle this problem, the Road Network Metric Learning for ETA (RNML-ETA) framework is introduced, which transfers knowledge from well-traveled routes to less-frequented ones through metric learning. This innovative deep learning approach effectively addresses the data sparsity challenge in road network modeling, outperforming existing models, particularly focusing on routes with limited available data.

GTRI (Georgia Tech Research Institute) is taking a cutting-edge approach to optimize aircraft maintenance scheduling and enhance risk management, both in military and commercial aviation sectors. They are harnessing the power of Artificial Intelligence (AI) and Machine Learning (ML) through a model-based systems engineering (MBSE) strategy. By leveraging big data analytics and multivariable optimization methods, GTRI is merging industry best practices with AI/ML to boost aircraft readiness while minimizing costs on a global scale. This initiative involves analyzing aircraft maintenance and routing data, using AI/ML techniques like random forest regression and classification to predict maintenance expenses and operational risks related to flight schedules. GTRI's MBSE logistics framework aims to dynamically optimize readiness costs and profit/loss scenarios in near-real time. The project promises to offer actionable insights to operations managers and planners, fostering innovation and experimentation in aerospace maintenance and logistics. GTRI's efforts align with the Joint AI Center's Joint Logistics National Mission Initiative and entail collaboration with industry partners like Delta Air Lines, with potential applications in Naval/Marine aviation fleets through collaboration with NAVAIR.