Experimental surgery performed by AI-driven surgical robot
- Thread starter JournalBot
- Start date
"An AI-controlled robot performed surgery on my hands. I give it three thumbs up!"
Upvote
83
(88
/
-5)
The silver lining here is that pranksters will have loads of fun doing prompt injection against AI plastic surgeons.
Source: https://smbc-comics.com/index.php?db=comics&id=1265#comic
Upvote
35
(39
/
-4)
How about fuck no?
Complications, malfunctions, lack of maintenance, hell fuck no.
Complications, malfunctions, lack of maintenance, hell fuck no.
Upvote
-8
(12
/
-20)
Given Intuitive have been sued multiple times over the past ten years for mistakes made when humans are operating the device it's probably not going to be much worse.
Upvote
21
(24
/
-3)
Upvote
-9
(15
/
-24)
I don't know what the popular "most recognized" examples are for AI, but it sounds like reinforcement learning, maybe not deep reinforcement learning since for medical cases, there usually isn't enough of a sample size to go too "deep".
In which case, historically, one would cite the models from Rich Sutton's lab or maybe DeepMind, as opposed to ChatGPT which is a LLM. But hey, ChatGPT is what the public recognizes, and I think they're trying to incorporate RL techniques into their newer "agentic" models so....
In which case, historically, one would cite the models from Rich Sutton's lab or maybe DeepMind, as opposed to ChatGPT which is a LLM. But hey, ChatGPT is what the public recognizes, and I think they're trying to incorporate RL techniques into their newer "agentic" models so....
Upvote
8
(10
/
-2)
A language model rooted around through some pig organs, that's not a surgeon.
Upvote
-15
(4
/
-19)
I placed a few dental implants using an "AI robot" and it was actually very impressive. I don't know what sensors it used, but essentially it knew where it was in relation to the patient based on the STL files from my surgical plan, and controlled orientation and depth while I performed the osteotomy. It's way more accurate and controlled than anybody doing it freehand. The problem as I see it is the procedure was actually half again as long as it normally takes me and according to the companies own literature, the accuracy is worse than 3d printing a surgical guide with a $300 resin printer and simple hand eye coordination and it cost like $150,000 and had software licenses of about a grand a month. Seems like a steep price to avoid spending 5 minutes and $10 3d printing and curing a surgical guide, and getting less accurate results at the end.
Upvote
76
(76
/
0)
ColdWetDog
Ars Legatus Legionis
The choice of doing cholecystectomies is interesting. It is actually a pretty easy procedure. Mostly. It does take some fairly delicate handling not to rip the vessels and ducts and, in humans, there is plenty of aberrant anatomy. Think twice, cut once.
Where it gets absolutely horrific is when there is a lot of inflammation. Like in cholecystitis. Like when you typically remove gall bladders. The swelling makes it hard to identify structures, makes structures much more fragile, makes everything bleed. Most surgeons would defer surgery until the bulk of the inflammation is gone - not always possible, but even after the acute event is over it can be pretty messy.
So this is an nice early demo. But we certainly are not at the point where you are going to just let these things loose in the hospital. I'm not really sure of the ultimate use case. It actually isn't all that hard to train a surgeon. You need time and patients. We have lots of them. You do need a training institution that is expensive but it isn't like these things are coming in cereal boxes. You'd still need the the building, nurses, anesthesia, techs and more techs. Sure, they won't get tired but everyone else might and most people just don't fail to boot up. As long as there is coffee around, anyway. You aren't likely to put them in the middle of nowhere as you are still going to need the techs, etc. and still have the economic issues that middles of nowhere tend to bring.
I would be interested in just how this company is going justify the economics.
Where it gets absolutely horrific is when there is a lot of inflammation. Like in cholecystitis. Like when you typically remove gall bladders. The swelling makes it hard to identify structures, makes structures much more fragile, makes everything bleed. Most surgeons would defer surgery until the bulk of the inflammation is gone - not always possible, but even after the acute event is over it can be pretty messy.
So this is an nice early demo. But we certainly are not at the point where you are going to just let these things loose in the hospital. I'm not really sure of the ultimate use case. It actually isn't all that hard to train a surgeon. You need time and patients. We have lots of them. You do need a training institution that is expensive but it isn't like these things are coming in cereal boxes. You'd still need the the building, nurses, anesthesia, techs and more techs. Sure, they won't get tired but everyone else might and most people just don't fail to boot up. As long as there is coffee around, anyway. You aren't likely to put them in the middle of nowhere as you are still going to need the techs, etc. and still have the economic issues that middles of nowhere tend to bring.
I would be interested in just how this company is going justify the economics.
Upvote
52
(53
/
-1)
Very curious how much it cost to train and operate. And how dependent it is on a reliable connection to the cloud. It would be pretty important to not lose your connection to the systems driving the robot.
It's pretty easy for a hospital to have power backup. Computing backup capable of running an LLM? Ummm....
It's pretty easy for a hospital to have power backup. Computing backup capable of running an LLM? Ummm....
Upvote
-2
(2
/
-4)
When the simple surgeries this can perform are no longer being done by humans, how will new surgeons gain the necessary experience to begin performing the more complicated surgeries the AI can't? How long until there's a real skill drain among upcoming surgeons?
If it's deployed correctly, that's an issue that can be avoided. But do you really trust tech bros and healthcare CEOs to deploy it correctly?
Upvote
26
(27
/
-1)
Nope, I don't trust the tech bros. On the other hand, there are plenty of smart people who are not capitalist sociopaths who could help bring helpful AI into being.
I also distrust reactionary hyperbole. (Not your post above, but a lot of posts on every AI article on Ars).
Upvote
-8
(7
/
-15)
This is certainly a step beyond where ML/AI is making inroads to radiographic diagnosis. What is nice about that application is the ability to completely human-check results before taking any action. Corrections during robotic surgery seem riskier than that, more reactive than preventive,
Upvote
4
(4
/
0)
They could be rivals.
After all, many people on the hit site LinkedIn list John Hopkins...
Come to think of it, John Hopkins, who played DS Dan Scott on the hit programme Midsomer Murders suddenly disappeared after series 8 so perhaps he's started a university.
Upvote
2
(3
/
-1)
Upvote
25
(25
/
0)
If it didn't, it would be all the more verisimilitude to replace a surgeon...
This is a horrifying idea. All it takes is one AI hallucination to ram a pair of scissors up through the liver and into the lung.
As a physician, the most important things I see AI doing in medicine is augmenting what we do, not replacing the thinking brain of an experienced physician.
A) Cover the busywork of charting. I spend half my time in the ED writing out why I did what I did, which takes away from patient care.
B) Cover the busywork of summarizing pt's relevant medical history from chart review. Patients almost NEVER give an accurate history, and those details they left out which were buried in a single outpatient chart buried amongst dozens from several years ago could give me information to prevent a life-threatening error.
C) Double checking our work. I never want AI to be the first pass on things because this will make physicians lazy and less skilled, but as a final check before I hit the "go" button it would be great to remind me to consider some rare diagnoses or double check my treatment plan. I think AI would be great to double check radiology reads as well.
Last edited:
Upvote
25
(27
/
-2)
That's the hump that has to be surmounted with any new tech though. I'd be surprised if the first 3d printers for surgical guides also didn't have a similar cost/benefit issue to overcome when compared to whatever previous methods were used
Upvote
2
(2
/
0)
I've seen many an Ars comment along those lines.
First it was how will junior artists learn skills to become experienced artists?
Then it was how will junior coders learn skills to become experienced coders?
Now you've brought up how will junior sturgeons learn skills to become experienced sturgeons?
The answer, as far as I can tell has been:
{this space intentionally left blank}
Let's pick all the low-hanging fruit and then call it a day appears to be the plan.
I fear junior comedy writers may be next.
Upvote
9
(11
/
-2)
Art and writing and coding are things you can perform on your own time to sharpen your skills. Doing the same with surgery is usually criminal.
Upvote
24
(24
/
0)
Upvote
2
(4
/
-2)
There is a similar robotic system for joint replacement. A key difference is it does not do the procedure. Instead, it acts as a digital jig. The surgeon uses cutting tools attached to the head and carves away bone like usual. The robot is datumed on the bone being worked on and limits extent of bone cutting. The final result are planes and holes that are precisely located and aligned for the implant to ensure the new joint is exactly where the surgeon decided it should go during surgical planning.
Upvote
7
(7
/
0)
Considering your usual posts and the ambiguous nature of this one, I'm going to assume people's sarcasm meter is pretty fucked for most folks today. Especially in light of the fact that robotic surgery is still pretty much an outlier even for humans operating it. It has high "potential", but the adoption and use is still slow - and that's among human operators, rather than AI.
There's not enough data about AI to be sure about their efficacy, and articles like this don't really lend themselves to an opinion one way or the other. This article is just reflective of the state of robotic surgery. "Promising". But given that it's AI, and the hype-machine about that shit is at full blast, I'd take anything said about it with a huge grain of salt.
Given your usual views, I read your post as sarcasm.
Sometimes that /s is pretty mandatory.
Upvote
5
(5
/
0)
I design surgical devices. Several years ago, a surgeon I work with explained why robotic surgeons will have a very difficult time ever getting cleared for use. It has nothing to do with technology. Surgeons make the rules for who can do surgery. For anyone else, it's illegal. Obviously they won't be in any hurry to permit surgical robots.
Ah, but but you might be thinking there's a loophole. What if the robot is working under the surgeons direction and supervision. The surgeon is the one doing the surgery on paper, but the robot is doing the actual work. Turns out something like this has happened and demonstrates what will happen.
Years ago, some surgeon had the bright idea of having meat robots (surgical assistants) do surgeries and he just signed off as them doing it under his supervision without actually being scrubbed in. He got slapped down hard!
There are likely countries where this monopoly isn't controlled by surgeons. Or maybe the government will override their objections. But it certainly creates a significant barrier most places.
Ah, but but you might be thinking there's a loophole. What if the robot is working under the surgeons direction and supervision. The surgeon is the one doing the surgery on paper, but the robot is doing the actual work. Turns out something like this has happened and demonstrates what will happen.
Years ago, some surgeon had the bright idea of having meat robots (surgical assistants) do surgeries and he just signed off as them doing it under his supervision without actually being scrubbed in. He got slapped down hard!
There are likely countries where this monopoly isn't controlled by surgeons. Or maybe the government will override their objections. But it certainly creates a significant barrier most places.
Upvote
8
(8
/
0)
The military-industrial complex verses the Surgeons will be a fun one to watch.
Upvote
2
(2
/
0)
ColdWetDog
Ars Legatus Legionis
Completely agree but the economics issue comes up hard and fast. If it is just backstopping a clinician, that isn't billable. Replacing a clinician is. Now, that is purely a consequence of our incredibly expensive, mostly backwards way of viewing medical costs but I don't think that is changing any time soon.
Sure, you can argue that fewer mistakes lead to fewer malpractice costs but malpractice really isn't a huge economic driver at present. You can argue that doing better medicine is cheaper but really, except for some low tech, well known preventative issues that hasn't been proven and is awfully hard to do so. So I'm interested in how they plan to sell this thing to hospitals.
It clearly is a machine that goes 'ping' and, as such, a hospital administrator is going to tend to look favorably at it but even in the NHS, I'm afraid that medical inflation has taken a big bite out of the pinging machine budget.
Upvote
1
(1
/
0)
Upvote
14
(14
/
0)