Researchers have developed a system to estimate external hand forces in manual material handling tasks using only RGB video and known box mass. The system uses a vision-language model that combines visual representations with text cues to predict the forces exerted on objects being handled. In an experiment, 35 healthy adults performed various lifting and carrying tasks while wearing no sensors or instruments. The results show that the system can estimate hand forces with reasonable accuracy, particularly when multiple cameras are used. This could lead to more efficient and scalable assessments of occupational physical exposure and injury risk. --- Why it matters: This matters to engineers and researchers in AI because it demonstrates a new approach to estimating continuous hand forces from visual data alone, which could be useful for various applications such as robotics, human-computer interaction, and biomechanical analysis. This technology has the potential to improve workplace safety and reduce the risk of injuries. Source: https://arxiv.org/abs/2608.22586