The reason nobody taught the robot what to do if it drops a needle is because nobody drops a needle. This is a cool thing, for sure, but all these AI miracles need to be taken with a grain of salt. What if the needle-drop breaks sterility? Does the robot really understand sterility or does it do what it's seen? There's no way to see every possible situation - you often have to improvise from a global understanding of what you're trying to accomplish, and what the constraints are. Robots don't understand the behavior of bacteria in relation to surgery, so it would be like teaching someone all the protocols for sterility without any of the underlying reasons. They would meticulously maintain a sterile field but if there was a surprise, as there so often is, they wouldn't know how to improvise a response from first principles. Without a world-view, machine intelligence will continue to be a little stupid. If you can't safely drive a car, can you do surgery? I know this is just basic stuff, tissue-handling and suture technique, but getting past the basic stuff is where machine intelligence tends to founder.
Johns Hopkins researchers have created a surgical robot that learns basic procedures simply by watching videos of human surgeons. Using AI similar to ChatGPT, the da Vinci system can now perform basic surgical tasks with human-level precision, marking a breakthrough in autonomous robotic surgery.