Scientists used recorded movements of people interacting closely with each other to develop a system that can generate more realistic animation of human interaction. This system could have many practical uses, such as improving video games or virtual reality experiences by making the interactions between characters feel more natural and realistic.
Read Full Paper →Researchers developed a new way to break down complex tasks into simpler steps, using computers to automatically identify these steps based on visual features learned from training data. This approach could have significant benefits for robots and other machines that need to perform complex tasks, as it allows them to learn and adapt more effectively without relying on human annotation or rule-based systems.
Read Paper →Scientists developed a new way of training artificial intelligence systems to behave safely while still performing well. This approach allows AI systems to learn safe behaviors from the start, which can help prevent accidents and lead to more reliable performance in real-world situations.
Read Paper →The researchers created a new system called RoboGhost that allows robots to move around in a more direct way by using language commands. This means the robot can understand what it needs to do based on a voice command, without having to go through many complicated steps first, which could help make robots more reliable and easier to control.
Read Paper →Scientists developed a new way to control robots that can walk on uneven terrain by breaking down their decision-making process into simpler, faster steps and slower, more thoughtful ones. This approach helped the robot succeed in tasks where other methods failed, which could lead to safer and more reliable use of humanoid robots in real-world environments.
Read Paper →