AI Equal With Human Experts in Medical Diagnosis, Study Finds (theguardian.com) 37
Artificial intelligence is on a par with human experts when it comes to making medical diagnoses based on images, a review has found. From a report: The potential for artificial intelligence in healthcare has caused excitement, with advocates saying it will ease the strain on resources, free up time for doctor-patient interactions and even aid the development of tailored treatment. Last month the government announced $305 million of funding for a new NHS artificial intelligence laboratory. However, experts have warned the latest findings are based on a small number of studies, since the field is littered with poor-quality research. One burgeoning application is the use of AI in interpreting medical images -- a field that relies on deep learning, a sophisticated form of machine learning in which a series of labelled images are fed into algorithms that pick out features within them and learn how to classify similar images. This approach has shown promise in diagnosis of diseases from cancers to eye conditions.
However questions remain about how such deep learning systems measure up to human skills. Now researchers say they have conducted the first comprehensive review of published studies on the issue, and found humans and machines are on a par. Prof Alastair Denniston, at the University Hospitals Birmingham NHS foundation trust and a co-author of the study, said the results were encouraging but the study was a reality check for some of the hype about AI. Dr Xiaoxuan Liu, the lead author of the study and from the same NHS trust, agreed. "There are a lot of headlines about AI outperforming humans, but our message is that it can at best be equivalent," she said.
However questions remain about how such deep learning systems measure up to human skills. Now researchers say they have conducted the first comprehensive review of published studies on the issue, and found humans and machines are on a par. Prof Alastair Denniston, at the University Hospitals Birmingham NHS foundation trust and a co-author of the study, said the results were encouraging but the study was a reality check for some of the hype about AI. Dr Xiaoxuan Liu, the lead author of the study and from the same NHS trust, agreed. "There are a lot of headlines about AI outperforming humans, but our message is that it can at best be equivalent," she said.
Good (Score:1)
The sooner we can de-symbolize the medical profession as being these infallible super-geniuses, the better. I have never dealt with a group of sloppier thinkers and workers than doctors. They are dangerous people and can change your life forever for the worse with one wave of their magic wand.
For a profession whose main symbol is the stethoscope, they sure don't seem to be able to listen. They are filled with biases, erroneous information, and logical fallacies.
Please, automate as much as possible.
Re: (Score:3)
Beating doctors at diagnoses is not hard. A nurse with a flowchart can do better. Of course, a doctor with a flowchart could do even better, but they are too pompous to use them.
Re:Good (Score:4, Informative)
Beating doctors at diagnoses is not hard. A nurse with a flowchart can do better. Of course, a doctor with a flowchart could do even better, but they are too pompous to use them.
Indeed.
This latest craze looks new to Millennials, but the "AI" of the late '80s could already beat doctors. They were called expert systems, and they were a goddamned flowchart. They reliably beat doctors in 1988. Doctors are terrifyingly bad, statistically. The subject they are attempting to specialize in is too big for the human mind to hold all at once but they refuse to accept this.
Those expert systems were carefully buried by the AMA, along with the studies that showed how much better they were than people. Several million people have died since then who would not have died if the expert system had been in use.
Re: (Score:2)
I'm usually to the right of conservative, but the US medical system is so corrupt and overpaid, its the one area where I call for a violently fascist takeover and socialism that would make Bernie scared.
Re: (Score:2)
"I prefer my doctor afraid of me and inches from bankruptcy."
When that act like that in the US, they send you to $20k worth of tests to be sure they didn't miss something and get sued. Even if insurance covers it, you are still paying for it one way or the other.
Re: (Score:2)
Not just tests--there's a lot of other things that happen due to defensive medicine. If things go wrong, people tend to sue the doctor--and even if it's a very stupid case (we're talking "The doctor had no control over what caused this" levels of stupid) it's still expensive to fight...and you still may lose, and that'll cascade down because now you can reasonably expect another, similar case to go the same way because it was 'proven' this time.
Note, too, that this is about AI being equal at best to humans
Re: (Score:2)
Buried? Quite conspiratorial of you.
No one figured out how to integrate them into routine practice in an automated way, properly.
The doctors ended up being swamped by alerts that were generated either wrongly or with inadequate information.
Earlier "A.I" was rule based and was applied to clean data. Doctors often had more patient context, beyond the data fed into the systems and ended up overriding them, and often with this frustration, correct recommendations as well.
Now doctors do use "algorithms", except
Re: (Score:2)
"The sooner we can de-symbolize the medical profession as being these infallible super-geniuses, "
Agreed, like all professions, 80% are crap in their job.
Lots of doctors are doctors just because their father was, a bit like the village idiot.
Re: (Score:2)
Re: (Score:2)
This is a solution in search of a problem. (Score:1)
Oh, wait....
Re: (Score:1)
To sell to governments who cant educate their own nations populations as doctors/experts, for use by NGO and charity groups.
A lab full of computers that works 24/7 for a third and fourth world nations.
The "money" is in the sale and support of such networked systems long term.
Not ready yet (Score:5, Insightful)
Considering how often doctors misdiagnose things I wouldn't really call this an impressive feat.
Re: (Score:2)
Considering how often doctors misdiagnose things even when they actually inspect the patient, having the AI at an equal level as the doctor when both are just using a picture is even less impressive than it sounds when you think they're comparing to the actual results the doctor gets in real life.
Re: (Score:1)
Would Machine Malpractice Cost the Same As Humans? (Score:4, Insightful)
As we've talked about Self-Driving Cars, Autonomous Flying Vehicles, machines have basically met the same level as humans but the expectation is that they must be perfect. There is the implicit requirement that machines cannot make mistakes, even if it is a mistake that a human would make.
But, we have a wrinkle here where a human doctor is protected against litigation when they make mistakes by malpractice insurance - if the AI is truly as good as a human doctor, would it's malpractice insurance cost just as much on a per-patient basis? I'm putting in that final qualifier because I would expect that the AI would be able to access test data faster and would be able to do more diagnoses in the same amount of time that a human doctor would take.
Are AIs equal in bedside manner? (Score:1)
While technical excellence is fantastic, there is a human element people still tend to expect. At least for now.
Re: (Score:1)
"Will an AI have any empathy? Any emotional reaction? "
LOL if you expect that from doctors. At best they present you with rehearsed theatrics. There's no more empathy there than when you bumped into a stranger and shot them a dirty look.
Re: (Score:3)
No. And when was the last time you got "empathy" from a doctor. If he cared that damn much about my feelings, she'd drop the fees a little bit.
Besides, due to the AMA, there won't be a Dr. Kiosk that you walk up to and get a diagnosis. You'll have to talk to the doctor, who will then type what you said into a program. Kind of like how they type what you said into Google now. They will be there to tell you to put one probe in your mouth and the other in your anus, or maybe the other way around depending
Re: (Score:2)
Will an AI have any empathy? Any emotional reaction?
Computer programs do what they are programmed to do.
If a computer is programmed to show empathy, then it will show empathy.
Human-level empathy is not hard to fake.
Doctor-level empathy is even easier.
Re: (Score:1)
Factor in the productive ability of a person if the computer grants approval for treatment.
An average fit younger citizen with good academic results? Treatment always approved.
Older person who is on a nations gov pension plan? Too old for treatment. Computer set to say nothing extra can be done. No approval.
Over 60 and the new treatment is set to never be approved. A decade later with a better per patient pr
Equal in only one way (Score:4, Insightful)
Equal at diagnosis. Currently. The AI will continue to improve, while it is a hit or miss situation with individual doctors. The improvements will be shared universally, which is hit or miss with individual doctors.
Not equal at focus. The computer will do one thing, and one thing only, and will be tireless at it. It will not be distracted by hunger, drowsiness, sexual desire or prejudice of the outcome (I don't think this patient has cancer, so I overlook the hard to see cancer).
Not equal on work time. The computer will even make a diagnosis while everyone else is asleep.
Not equal on cost. Hardware is cheap. More so when compared to a doctor's salary.
Does it support a Wallet Biopsy? (Score:3)
I'm not even sure if the doctor knew they were doing it. Implicit bias are real; and a bitch.
LOL, no, it's not. (Score:1)
Re: (Score:2)
"you can't trust it for something as important as this"
As what? Determining that you broke a leg? Have cancer? A cold?
Sure, some diagnosis may always be so consequential to require human analysis to be sure its accurate, but to test for athlete's foot may be fine with a computer. We already trust nurse practitioners and regular nurses for some ailments. This would just be an extension of that.
Re: (Score:1)
Re: (Score:2)
"Thinking" is a general-purpose tool. Because people can think, doctors can learn how to diagnose illnesses. Mathematicians can learn how to do math. These are not jobs that people are especially good at, without specialized training.
Medical AI, on the other hand, is specifically tailored to diagnose illnesses. It does so because that is what it is designed to do. Just as a calculator is better than a human at math, a computer will soon be far better at diagnosing illnesses than humans.
State boards will try to nix this (Score:2)
The function of state medical boards is to keep medical procedures as old-fashioned and lucrative for as long as possible, so expect them to maniacally oppose this. But what if beside every Indian casino there were a medical building staffed by personnel who could competitively apply these AIs to your problems?
Time to replace doctors! (Score:2)
Hammers are better than humans too (Score:2)
The trick isn't to make a machine that's better than a human at performing a given task. That's never been at all difficult. Hammers hit harder than hands. Screw drivers turn better than fingers. Drills, pens, buttons, anvils, hoists, cars, boats, desks.
Just about any machine we've ever built does some task better than humans do. That's why we build them. That's what a tool is.
What makes each and every one of these machines no where near as good as a human is that it needs its environment to be perfect
Re: Hammers are better than humans too (Score:2)
Wrong focus (Score:1)
Re: (Score:2)
use
to make a new line on /.