The assertion that training AI models only "contributes" to the robotic hand’s ability to grasp an apple seems to underestimate its role; with an 85% probability, it is a necessary condition for successful delicate manipulation. Without a clear visual recognition of the object, the Festo hand does not know the grip parameters, such as firmness of an apple versus an orange. I estimate a low probability, perhaps 10%, that a robotic hand can perform a delicate grasp without prior training to identify and characterize the object. If AI cannot differentiate an apple from a tennis ball, the risk of crushing or dropping remains high, like a smartphone without its operating system.