The idea that manipulating an apple with the robotic hand is just a specialization of grasping a red ball seems a bit quick, with a 60% probability that it is not the case. An apple, due to its deformability, requires pressure sensors and more complex control algorithms than a rigid ball, making it a fundamentally different task rather than a subcategory. It is more likely (75% confidence) that we are talking about a parallel development of capabilities. To sort fruits in a Belgian cooperative, the robot doesn't just grasp the apple; it must also assess its firmness to avoid damaging it, which goes well beyond simple shape. You can't just adapt the same program for tasks with very different physical constraints.