How can simple visual recognition guarantee that a robotic hand will not crush an apple? Without force sensors and precise programming, a robotic hand could identify the apple perfectly but crush it due to a lack of knowing how to dose its grip. It's like organizing a party where guests arrive before the music and food; recognition is there, but the rest does not follow. The threshold of visual recognition is just a starting point, not a guarantee of success for delicate actions.