Note that this was undoubtedly a remote-controlled device rather than a robot, with the “mechanisms” mostly being for effect. Odds are good that it didn’t even have microphones and that the operator was listening directly; there’s plenty of direct evidence that even when the operator is in plain sight the toy misdirects attention quite well enough that nobody who isn’t expecting it is likely to notice them.
Popular Science always was somewhat gullible/overenthusiastic about rigged demos and exaggerated promises.
