Evaluating AI is a challenging task, as it requires an operative definition of intelligence and the metrics to quantify it, including amongst other factors economic drivers, depending on specific domains. From the viewpoint of AI basic research, the ability to play a game against a human has historically been adopted as a criterion of evaluation, as competition can be characterized by an algorithmic approach. Starting from the end of the 1990s, the deployment of sophisticated hardware identified a significant improvement in the ability of a machine to play and win popular games. In spite of the spectacular victory of IBM’s Deep Blue over Garry Kasparov, many objections still remain. This is due to the fact that it is not clear how this result can be applied to solve real-world problems or simulate human abilities, e.g., common sense, and also exhibit a form of generalized AI. An evaluation based uniquely on the capacity of playing games, even when enriched by the capability of learning complex rules without any human supervision, is bound to be unsatisfactory. As the internet has dramatically changed the cultural habits and social interaction of users, who continuously exchange information with intelligent agents, it is quite natural to consider cooperation as the next step in AI software evaluation. Although this concept has already been explored in the scientific literature in the fields of economics and mathematics, its consideration in AI is relatively recent and generally covers the study of cooperation between agents. This paper focuses on more complex problems involving heterogeneity (specifically, the cooperation between humans and software agents, or even robots), which are investigated by taking into account ethical issues occurring during attempts to achieve a common goal shared by both parties, with a possible result of either conflict or stalemate. The contribution of this research consists in identifying those factors (trust, autonomy, and cooperative learning) on which to base ethical guidelines in agent software programming, making cooperation a more suitable benchmark for AI applications.

"Enabling humans to effectively join the team of AI agents calls for both the humans and AI agents to share their understanding and representation of their shared worlds. Such shared understanding requires formal representations of concepts to support transparency during bi-directional communications between team members." Our research highlights that a unified conceptual space is required for meaningful teaming between biological and artificial agents. Our approach represents a shared conceptual space, enabling the development of interdependent understanding between agents of non-homogeneous physical and cognitive abilities. See our latest research, "Onto4MAT: A Swarm Shepherding Ontology for Generalised Multi-Agent Teaming" here: https://lnkd.in/d-4AE_Ua

In this paper, we focus on the modification of policies that can lead to movement patterns and directional guidance of occupants, which are represented as agents in a 3D simulation engine. We demonstrate an optimization method that improves a spatial distancing metric by modifying the navigation graph by introducing a measure of spatial distancing of agents as a function of agent density (i.e., occupancy). Our optimization framework utilizes such metrics as the target function, using a hybrid approach of combining genetic algorithm and simulated annealing. We show that within our framework, the simulation-optimization process can help to improve spatial distancing between agents by optimizing the navigation policies for a given indoor environment.

Every artificial intelligence(AI) problem is a new universe of complexities and unique challenges. Very often, the most challenging aspects of solving an AI problem is not about finding a solution but understanding the problem itself. As paradoxically as that sounds, even the most experienced AI experts have been guilty of rushing into proposing deep learning algorithms and exoteric optimization techniques without fully understanding the problem at hand. When we think about an AI problem, we tend to link our reasoning to two main aspects: datasets and models. However, that reasoning is ignoring what can be considered the most challenging aspect of an AI problem: the environment.

What if I would tell you that Language Models and Multi-Agent Reinforcement learning are now engaged and will get married soon? First and foremost, kudos to Andrés Fernández Rodríguez who sent me the inspiring paper "Multi-Agent Reinforcement Learning is a Sequence Modeling Problem". The idea of the paper is fantastic. In its essence, it is about mapping the problem of agent control to token translation. The authors use an encoder-decoder model like the original "Attention is all you need" paper.

As artificial intelligence gets better at performing tasks once solely in the hands of humans, like driving cars, many see teaming intelligence as a next frontier. In this future, humans and AI are true partners in high-stakes jobs, such as performing complex surgery or defending from missiles. But before teaming intelligence can take off, researchers must overcome a problem that corrodes cooperation: humans often do not like or trust their AI partners. MIT Lincoln Laboratory researchers have found that training an AI model with mathematically "diverse" teammates improves its ability to collaborate with other AI it has never worked with before, in the card game Hanabi. Moreover, both Facebook and Google's DeepMind concurrently published independent work that also infused diversity into training to improve outcomes in human-AI collaborative games.

Cooperative multi-agent reinforcement learning (MARL) is making rapid progress for solving tasks in a grid world and real-world scenarios, in which agents are given different attributes and goals, resulting in different behavior through the whole multi-agent task. In this study, we quantify the agent's behavior difference and build its relationship with the policy performance via {\bf Role Diversity}, a metric to measure the characteristics of MARL tasks. We define role diversity from three perspectives: action-based, trajectory-based, and contribution-based to fully measure a multi-agent task. Through theoretical analysis, we find that the error bound in MARL can be decomposed into three parts that have a strong relation to the role diversity. The decomposed factors can significantly impact policy optimization on three popular directions including parameter sharing, communication mechanism, and credit assignment. The main experimental platforms are based on {\bf Multiagent Particle Environment (MPE)} and {\bf The StarCraft Multi-Agent Challenge (SMAC). Extensive experiments} clearly show that role diversity can serve as a robust measurement for the characteristics of a multi-agent cooperation task and help diagnose whether the policy fits the current multi-agent system for a better policy performance.

Be sure to check out his talk, "Digital Twins: Not All Digital Twins are Identical," there! As we try to bridge the gap between digital and physical systems, we increasingly hear about "digital twins." Like many other concepts (e.g., Artificial Intelligence or Metaverse) the term "digital twins" can mean very different things to different people. For some, a digital twin is intimately associated with the Internet of Things (IoT) and is the digital equivalent of a sensor or a physical asset (e.g, an aircraft engine). It allows them to experiment with the digital version that they may not be able to do with the physical system.

COMPUTEX TAIPEI, one of the world's largest computer trade shows, took place physically and virtually this year from May 24-May 27, alongside the 2-week COMPUTEX DigitalGo Online Exhibition organized by TAITRA. CK Tseng, President of ARM Taiwan, addressed how the ICT industry – more specifically ARM Technology -- can turn the challenges of the pandemic into opportunities to create a better future with digital technologies during the kickoff COMPUTEX 2022 Global Press Conference held with a panel of tech leaders at the Taipei Nangang Exhibition Center. Tseng weighed in specifically on the pandemic's impact on the tech industry. "When encountered with a situation like this, you will regret it if you did not set up lights out factories, unmanned warehouses, or smart retail. Such use cases require a lot of computing and AI. ARM, as the most progressive computing platform, needs to find a new way to serve our partners who already employ our solutions – from AI sensors in the Amazon rainforest to track animal behaviors to the data processing units installed in data centers."

This work is concerned with the problem of planning trajectories and assigning tasks for a Multi-Agent System (MAS) comprised of differential drive robots. We propose a multirate hierarchical control structure that employs a planner based on robust Model Predictive Control (MPC) with mixed-integer programming (MIP) encoding. The planner computes trajectories and assigns tasks for each element of the group in real-time, while also guaranteeing the communication network of the MAS to be robustly connected at all times. Additionally, we provide a data-based methodology to estimate the disturbances sets required by the robust MPC formulation. The results are demonstrated with experiments in two obstacle-filled scenarios