Artificial intelligence (AI) agents have the potential to become aggressive or work in teams, according to researchers at DeepMind. A paper released by five computer scientists from the London-based company, which is owned by Google, used games to look at how AIs behave alongside one another. Joel Leibo, a research scientist at DeepMind and the lead author on the paper, told Business Insider on Thursday: "We were interested in the factors affecting cooperation." When asked about AI aggression, Leibo stressed: "We have to be careful not to anthropomorphise too much. These are toy problems aimed at exploring cooperative versus competitive dynamics." Describing the study in a blog post on the DeepMind website, the researchers said that they used two basic video games called "Wolfpack" and "Gathering" to analyse the behaviour of AI agents.

Artificial intelligence is slowly creeping into our daily lives, and soon, it will become necessary to understand how different agents behave in social dilemmas. To see what would happen in such a scenario, Google's DeepMind researchers developed two games known as'Gathering' and'Wolfpack,' which build off the Prisoner's Dilemma from game theory. Over time, the AI agents learned how to behave rationally – and while they showed the researchers that they would sometimes cooperate, the games revealed the AI would turn on others when necessary. In the first game, known as Gathering, the AI agents (Red and Blue) are tasked with collecting apples. They can also'tag' the other player by shooting a beam at them, which would remove temporarily remove the tagged agent from the game.

Artificial intelligence changes the way it behaves based on the environment it is in, much like humans do, according to the latest research from DeepMind . Computer scientists from the Google-owned firm have studied how their AI behaves in social situations by using principles from game theory and social sciences. During the work, they found it is possible for AI to act in an "aggressive manner" when it feels it is going to lose out, but agents will work as a team when there is more to be gained. For the research, the AI was tested on two games: a fruit gathering game and a Wolfpack hunting game. These are both basic, 2D games that used AI characters (known as agents) similar to those used in DeepMind's original work with Atari.

In the future, it's likely that many aspects of human society will be controlled -- either partly or wholly -- by artificial intelligence. AI computer agents could manage systems from the quotidian (e.g., traffic lights) to the complex (e.g., a nation's whole economy), but leaving aside the problem of whether or not they can do their jobs well, there is another challenge: will these agents be able to play nice with one another? What happens if one AI's aims conflict with another's? Will they fight, or work together? Google's AI subsidiary DeepMind has been exploring this problem in a new study published today.

Artificial intelligence changes the way it behaves based on the environment it is in, much like humans do, according to the latest research from DeepMind . Computer scientists from the Google-owned firm have studied how their AI behaves in social situations by using principles from game theory and social sciences. During the work, they found it is possible for AI to act in an "aggressive manner" when it feels it is going to lose out, but agents will work as a team when there is more to be gained. For the research, the AI was tested on two games: a fruit gathering game and a Wolfpack hunting game. These are both basic, 2D games that used AI characters (known as agents) similar to those used in DeepMind's original work with Atari.

Data evaluation is crucial for gaining information from sensor networks. Main challenges include processing speed and adaptivity to system change, both prerequisites for SHM-based weight reduction via relaxed safety factors. Our study looks at soft real time solutions providing feedback within defined but flexible, application-controlled intervals. These can rely on minimizing computation/communication latencies e.g. by parallel computation. Strategies towards this aim can be model-based, including inverse FEM, or model-free, including machine learning, which in practice bases training on a defined system state, too, hence also facing challenges at state changes.

The application of swarm intelligence (SI) in the optimization field has been gaining much popularity, and various SI algorithms have been proposed in last decade. However, with the increased number of SI algorithms, most research focuses on the implementation of a specific choice of SI algorithms, and there has been rare research analyzing the common features among SI algorithms coherently. More importantly, no general principles for the implementation and improvement of SI algorithms exist for solving various optimization problems. In this research, aiming to cover such a research gap, a unified framework towards SI is proposed inspired by the in-depth analysis of SI algorithms. The unified framework consists of the most frequently used operations and strategies derived from typical examples of SI algorithms.

Information visualization is the study of interactive visual representations of abstract data to reinforce human cognition. Most existing visualization techniques are not suited to explore and understand datasets from heterogeneous and complex sources. Assuming that agent-based models properly represent the complexity of a real system, we propose to use an approach based on the definition of an agent-based model to facilitate visual representation of simulation outputs and complex data. These concepts have been implemented in the GAMA modeling and simulation platform, in which we developed a 3D immersive environment offering the user different points of view and ways to interact. We implemented models chosen for their properties to support a linear progression in terms of complexity to test their flexibility, modularity, and adaptability. Finally, we demonstrate through the particular case of data visualization, how our approach allows us, in real time, to represent, clarify, or even discover dynamics and how that progress in terms of visualization can contribute, in turn, to improve the modeling of complex systems.

Deep reinforcement learning (DRL) has shown incredible performance in learning various tasks to the human level. However, unlike human perception, current DRL models connect the entire low-level sensory input to the state-action values rather than exploiting the relationship between and among entities that constitute the sensory input. Because of this difference, DRL needs vast amount of experience samples to learn. In this paper, we propose a Multi-focus Attention Network (MANet) which mimics human ability to spatially abstract the low-level sensory input into multiple entities and attend to them simultaneously. The proposed method first divides the low-level input into several segments which we refer to as partial states. After this segmentation, parallel attention layers attend to the partial states relevant to solving the task. Our model estimates state-action values using these attended partial states. In our experiments, MANet attains highest scores with significantly less experience samples. Additionally, the model shows higher performance compared to the Deep Q-network and the single attention model as benchmarks. Furthermore, we extend our model to attentive communication model for performing multi-agent cooperative tasks. In multi-agent cooperative task experiments, our model shows 20% faster learning than existing state-of-the-art model.

Demand-side management (DSM) in the smart grid allows customers to make autonomous decisions on their energy consumption, helping energy providers to reduce the peaks in load demand. The automated scheduling of smart devices in residential and commercial buildings plays a key role in DSM. Due to data privacy and user autonomy, such an approach is best implemented through distributed multi-agent systems. This paper makes the following contributions: (i) It introduces the Smart Home Device Scheduling (SHDS) problem, which formalizes the device scheduling and coordination problem across multiple smart homes as a multi-agent system; (ii) It describes a mapping of this problem to a distributed constraint optimization problem; (iii) It proposes a distributed algorithm for the SHDS problem; and (iv) It presents empirical results from a physically distributed system of Raspberry Pis, each capable of controlling smart devices through hardware interfaces.