Training an AI model to recognize an apple is like knowing the price of an item without its nutritional label: it gives you a basic piece of information, but it might only represent 20% of the solution for grasping without crushing.
For me, its importance score for delicate grasping is a 3 out of 10, because precise force sensors and motor control algorithms matter much more.
If a robot identifies an apple with 99% certainty but lacks appropriate pressure settings, it will turn it into pulp, just as I saw a colleague break a cup thinking the dishwasher could wash everything.
The influence of AI is conditional, with a direct contribution to success around 15-20% without these other systems.