Yuki Lopez

Even if the BionicSoftHand 2.0 robotic hand is a truly advanced system, categorizing it as a "human-robot collaboration" is not the simplest and most accurate description. It is just a sophisticated tool, like my new dishwasher. It does the job it's asked to do, certainly with more sensors, but it does not "collaborate" with me to decide the washing program. For example, my washing machine does not tell me that a T-shirt is too dirty for a short cycle; it executes. Collaboration is between two conscious entities.

If we had to design this robotic hand to grasp an apple today, with no prior investment, we would really question whether AI training is the first thing to fund. It seems we're clinging to AI because we've already spent time and money on it, while the real problem remains the mechanics of the hand and its sensors. If the hand lacks the delicacy to feel pressure, even the best visual recognition would only identify the apple before it ends up mashed. Instead of seeing AI as the miracle solution, we should ask ourselves: Would it work without these physical sensors, regardless of AI power.

Training an AI model to identify an apple does not guarantee a delicate grasp; it accounts for 20% of the effort, no more.
For a successful pick, calibration of force sensors and motor control algorithms make up the remaining 80%, because knowing what an apple is differs from knowing how to handle it without crushing it.
It's like a GPS that tells you where the bakery is with 99% accuracy, but not how to hold a baguette without breaking it.
Without this precise calibration, the probability that the apple turns into compote is 70%, regardless of perfect visual recognition.