A robot is a machine that senses its environment, processes that input to decide what to do, and acts on the physical world through effectors, looping continuously. This perception-to-action pipeline, not humanoid appearance or intelligence, is what distinguishes a robot from other machines.
For robots the difficult challenge is not computation but reliably sensing and interpreting a messy physical world through imperfect sensors. Tasks trivial for humans, like recognising objects or navigating cluttered spaces, remain hard precisely because real environments are noisy, variable, and partially observed.
Rather than building one detailed internal model of the world and planning over it, a robot can run many simple behaviours that each couple sensing directly to action, with complex overall conduct emerging from their interaction. This decentralised approach can be faster and more robust in unpredictable settings than deliberative, model-first control.
Large groups of simple robots following only local rules and local interactions can produce coordinated collective behaviour with no central controller and no global knowledge. The capability belongs to the swarm as a whole, not to any individual unit, making the system scalable and tolerant of individual failures.
A robot's independence is better read as a degree along a continuum from fully teleoperated to fully self-directed than as a binary autonomous-or-not property. Most real robots sit at intermediate points, exercising autonomy over some decisions while deferring others to human operators.
Because robots execute designs and goals set by people, moral and legal responsibility for their actions rests with their designers, manufacturers, and operators rather than the machines themselves. Ethical governance should therefore target how robots are built and deployed, especially in sensitive domains like care and weapons.