Hacker News Reader

Comment by alexholehouse | Hacker News Reader

So I was having a debate with a colleague a few weeks ago: the suggestion was that in 20 years time the a large percentage of surgeries done in hospitals will be done by robots .

Importantly, we didn't specify which surgeries, but through the conversation the implication was not that robots would be removing millions of weird moles and skin marks - we're talking serious, long surgeries.

I'd be really interested to hear the HN community's view on this. I have my own (strong) opinions based on my knowledge of AI and ML as a field, but perhaps other people have their own strong opinions (and perhaps people can change my mind!)

View on HN

I'm having my kidney removed for transplant in February. The hospital offered to have the surgery performed with robotic tools where I'm never actually touched by surgeons -- they are controlling the tools from a workstation.

These tools have come under fire because a certain number of surgeries have gone wrong using them, but it seems like when you compare the number of incidents vs the number of surgeries performed the percentage of error is significantly smaller than those performed by human hands.

That said, I definitely agree that it is only a matter of time before a robot will be handling the surgery from start to finish, having learned how to perform the task by trained surgeons. The complications arise when the robot runs into difficult situations and things that happen outside of their defined work patterns.

I'm founder and lead developer of a ML startup where we trained our system to do realistic 3D modeling for human 3D animation. From my perspective, I think it will be less than 20 years before the vast majority of longer surgeries will be automated. We already have robotic assistance, and remote robotic surgical hands for when the doctor and patient are in separate locations. There is too much money to be made for this to remain non-main stream.

If you think that liability issues pose a significant barrier to entry for self-driving cars getting to market, just wait until you consider robotically automated surgeries.

Just wait until the probability of error with a robot surgeon is lower than that of a human surgeon.

Even if it's lower, a single mistake from a robot surgeon will not be forgiven. That's the same problem with self-driving cars. I'm pretty sure that they're already safer than a human driver, but a single mortal accident for a technical problem will be all over the news and block any progress in legislation.

Driving is different. Thought `robot is better driver than me` is much harder to accept than `robot is a better surgeon than someone i don't know`.

You have to specify what you mean by "done by robots". Will a qualified surgeon still need to be present for the operation to take place?

Probably, but they may be alone supervising 3 operations in parallel rather than having a team of junior surgeons and nurses to help them on one operation.

An intermediate step would be robotic tools that could be controlled by a surgeon as effectively as their own hands.

As such, I do not believe we will see majority-robotic surgery within 20 years.

"An intermediate step would be robotic tools that could be controlled by a surgeon as effectively as their own hands."

But that is basically already were we are today right? And it's way more efficient than "as their own hands": a friend surgeon was explaining me that robotic tools were used not only because they were way more precise than human hands but also because they allowed for far greater freedom of movements, being able to rotate on way more axis than our human hands.

This is one of the main "problem" with robots: it's not just that they're "as good" as we are. It's that they're way more precise and can do things we can't.

There are also, if I'm not mistaken, surgeries that have been performed across countries, with the surgeon in one country controlling the robotic arms in another country.

I agree with your assessment of the situation, but not with the conclusion that robotic tools are as effective as hands:

Robotic tools have been optimised for specific, planned and clearly-defined surgeries, and I know they can offer less invasive keyhole surgeries.

But, they do not nearly emulate hands in a general way - much bigger breakthroughs are required for robotic tools to be useful in emergency surgery.

  > we're talking serious, long surgeries

well, 20 years is a long time, so it is hard so speculate. But I'll say, probably not.

I'm a beginner at the math of AI/ML, but I can say this: almost every task we've ever assumed requires human general intelligence - short of the old "world domination" standby - has turned out to be entirely doable by a sufficiently specialized AI/ML algorithm. Even spoken conversation is turning out more amenable than originally proposed, once you confine the software to one kind of conversation.

Reply on news.ycombinator.com