What are modern robots?
A modern robot combines a mechanical body, sensors, actuators, compute and control software. Some systems operate in structured environments such as production lines, while humanoid and mobile robots attempt to work in more varied spaces. Performance therefore depends on perception, control, safety, reliability and mechanical engineering as much as it depends on an AI model.
Humanoid and industrial robots
Industrial robots are typically optimized for repetitive and precise tasks. Humanoids aim to take advantage of environments originally designed for people, including stairs, doors and human-scale tools. A human shape does not automatically imply greater intelligence, so useful evaluation focuses on sustained walking, manipulation, task completion, recovery from errors and operation outside carefully staged demonstrations.
Physical AI, perception and control
Physical AI connects learning and perception with decisions in the real world. Cameras, depth sensors, force sensing and tactile input help estimate the environment, while planning and control algorithms translate those estimates into motion. Multimodal models can add language and visual understanding, but they do not replace the fast, safety-critical control loops required for physical machines.
How should robot capability be measured?
Visual demonstrations are informative but incomplete. More useful measures include task success rate, completion time, repeatability, battery life, payload, precision, human intervention rate and performance outside the laboratory. Commercial deployments also depend on operating cost, maintenance, safety certification and integration with existing systems.
Trends worth following
Fast-moving areas include humanoids for factories and services, warehouse mobile robots, autonomous drones and vehicles, vision-language-action models, simulation for policy training and edge computing. Cybersecurity is also becoming more important as robots connect to cloud services, remote management systems and AI agents.
This is an evergreen Mikhbar editorial guide, updated when the underlying concepts or technologies materially change.