The assertion that grasping an apple is just a direct specialization of grasping a ball seems to have a 60% probability of being too simplistic.
Our priority should be that an apple, with its irregular shape and variable firmness (especially if it is a bit soft, as sometimes found at the market), requires much more nuanced sensor control.
The likelihood that a robot manages an apple bruised during transport optimally is low, much more complex than handling a rigid ball.
For example, a system trained for uniform balls could damage a fragile apple if pressure sensors are not recalibrated for the specific properties of fruits, a necessary update for real agricultural applications.