How Should a Robot Show That It’s Still Trying to See?

A robot that looks twice is better than one that guesses. But the person beside it needs to know which state it is in. Focus Taiwan watched Solomon’s humanoid take spoken or written instructions, locate an object about five meters away, walk over, and grasp it. Solomon says its “active perception” runs a loop of searching, selecting, zooming, and evaluating instead of trusting one camera frame. The public demo is encouraging; it is not field evidence, and neither source gives false-target rates or shows what nearby workers see. Before the robot moves, I’d want one plain signal visible from the floor: searching, target found, or unsure. If it changes its viewing angle or rolls closer just to get a better look, that should look different from committing to the pick. Show the selected object too—not as a tiny box on an engineer’s laptop, but on a nearby screen or projected marker a coworker can correct. A robot pausing to look again is good behavior. How should it make that pause legible without turning every factory into a wall of warning lights?