I'm working for a small company that is trying to bridge the gap and make something good that can run on the consumer VR/AR hardware that's coming out in the next year or three. Lots of interesting problems to solve.
https://www.intuitive.com/en-us/products-and-services/da-vin...
In my experience in a related field, simulators are about 2% as useful as advertised. The cousin thread explains how difficult building a flesh simulator is, so I wouldn't expect surgical experience on a simulator to be very useful.
In the case of robotic surgery simulator, you are using the controls to control the arm, but to interact with what? If you just want to move the arm around and maybe interact with some rigid objects sure that's easy. Would that add any training value to the surgeon? Probably not. You can get value only when the simulator includes a simulation of something the surgeon would have to face eventually - organic mass of the human body. Simulating that is hard, and I doubt anyone would invest much into it when you can train surgeons on alternative physical objects like pigs.
To run a high-fidelity simulator of human body that is useful for surgeons, you need a LOT more. And I doubt it can be data-driven. Data-driven simulators for things like autonomous cars are just coming up, and these are way simple as the agent (the autonomous car) doesn't directly change the environment when it's driving around (as in, you don't have to simulate the car colliding into a traffic pole and then breaking it, etc.)
To simulate a human body, you need to be able to capture the material properties of different layers, some of which are fluids, and then also the interaction between them and how different organs react to a surgical operation. It's a very hard problem.
Why do all that when you have animal corpses readily available? There's no need to create a problem that doesn't exist.
a) thousands of hours is far far far too little data. This is a total ballpark estimate, with only knowledge of the ML and basically zero knowledge of the surgery side, but I would expect three to six orders of magnitude greater, i.e. millions to billions of hours necessary to train a machine learning model to do something like that, to a standard.
b) the big problem is not the procedures themselves, but edge cases and things going a little bit wrong.
c) you are fine with tiny flukes in your generated images, but you won't be very happy with an ML model whose tiny flukes lead to internal bleeding.
d) unfortunately (with contemporary models) we can't easily explore latent space, which might possibly contain regions corresponding to catastrophic robot arm movements. the chance isn't that high, but it's not a chance most'd be willing to take at this point.
On the other hand, simulating say skin/muscle/blood or anything that closely resembles human body is near impossible. Without that, a simulator is pretty much useless and it'll probably easier to train the surgeons on real robot + some animal like pig