Being trained on videos means it has no ability to adapt, improvise, or use knowledge during the surgery.

Edit: However, in the context of this particular robot, it does seem that additional input was given and other training was added in order for it to expand beyond what it was taught through the videos. As the study noted, the surgeries were performed with 100% accuracy. So in this case, I personally don't have any problems.

Have you considered that the machine is made by a collection of humans?

Imagine if the Tesla autopilot without lidar that crashed into things and drove on the sidewalk was actually a scalpel navigating your spleen.

Theres no evidence they will ever reach quality output with infinite data, either. In that case, quality matters.

I actually don't think that's the problem, the problem is that the AI only factors for visible surface level information.

AI don't have object permanence, once something is out of sight it does not exist.

If you read how they programmed this robot, it seems that it can anticipate things like that. Also keep in mind that this is only designed to do one type of surgery.

I'm cautiously optimist.

I'd still expect human supervision, though.

No we don't know. We are not AI researchers after all. Nonetheless I'm more inclined to defer to experts then you. No offence, (I mean there is some offence, because this is a stupid conversation) but you have no qualifications.

Absolutely stupid example because that kind of assumes medical professionals have the same standard as Elon Musk.

You assume an Antarctic research facility lacks funds?

Elon Musk literally owns a medical equipment company that puts chips in peoples brains, nothing is sacred unless we protect it.

Yes. But if the machine has proven to work reliably it will usually do so for its lifetime, while humans are prone to e multitude of errors. Especially in the medical field.

It's less of an unknown and more of a "it has never demonstrated any such capability."

Btw both OpenAI and Deepmind wrote papers proving their then models would never approach human error rate with infinite training. It correctly predicted performance of ChatGPT4.

Unless the videos have proper depth maps and identifiers for objects and actions they're not going to be as effective as, say, robot arm surgery data, or vr captured movement and tracking. You're basically adding a layer to the learning to first process the video correctly into something usable and then learn from that. Not very efficient and highly dependant on cameras and angles.

Into volunteers it's not standard practise to randomly put a chip in your head.