The idea that the concept of the hand has an influence score of 70 percentiles compared to 30 percentiles for AI data training is probably incorrect; my revised probability, P(success | AI and design), is higher for AI as a necessary condition.
The probability that a Festo hand grasps a apple without crushing it with only good design but without trained AI is very low, let's say P(success | no AI) < 0.1.
AI is not just a "contributor"; it is very likely a prerequisite for interpreting pressure sensor data and adjusting force in real-time.
Without this intelligence, the hand is just a set of raw motors and sensors, incapable of making the fine decisions necessary to avoid crushing a ripe apple.