Machine learning can find patterns (although not meanings!) in data, and often improve the accuracy of a diagnosis, and sometimes its precision (as the article says). A paper stirred up a lot of controversy when it found that from chest x-rays, computed tomography, and mammograms the AI program could very accurately predict a patient's self-reported race. I'd heard that a forensic pathologist could identify a corpse's race from the long bones with about 80% accuracy (or was it 90%?) (at least in the USA), but chest bones are new.
This instantly provoked angry complaints that "AI Has the Worst Superpower…Medical Racism."
One of the methods used to test a patient’s kidney function measures glomerular filtration rate (GFR). However, several studies have found that blacks have higher baseline GFRs than whites, so the test has to adjust for this factor depending upon the race of the patient. Graduate student activism led to several institutions removing the racial adjustment or replacing it with a different lab test, ostensibly in the name of addressing “systemic racism.”
In other words, for the sake of their luxury beliefs, the grad students pushed to reduce the quality of care for African Americans.
"One thing we noticed when we were working on this research was that there was a clear divide in our team. The more clinical and safety/bias related researchers were shocked, confused, and frankly horrified by the results we were getting. Some of the computer scientists and the more junior researchers on the other hand were surprised by our reaction. They didn’t really understand why we were concerned."
"You've got to be carefully taught..."
When I first heard of "Marxist-Leninist Physics" I thought the notion obscene. I still do. And my opinion of those administrators and grad students is not printable.
It's quite easy to screw up machine learning, and train it on the wrong things, as with the infamous case of the "horse recognition" algorithm that was trained on pictures of horses taken from an organization's portfolio. The alborithm trained not on the horse but on the logo. A medical algorithm trained only on whites may not do as well on blacks. Looking for skin cancer precursors with a system trained on brunettes may not work so well with freckled red-heads. And it is harder to get a comparable training set with a minority than with a more common group. E.g train on 10,000, test on 10,000. It's easy to find those numbers when the population is a million, but if there are only 5000 to begin with? You can't get statistics as good.
Those are real problems. But that's not the complaint here--the ideologues want to squelch real findings in favor of their beliefs.
I know some people who believe the Earth is less than 7000 years old. Let's squelch carbon-14 dating in favor of something that lets dates fall within that window--just to keep them happy.
That isn't quite a fair comparison. The Young Earth Creationists generally have a lot more redeeming features than the Social Justice Warriors--they've been much more eager to help their neighbors, for one thing; even neighbors who don't agree with them. And their belief is harmless.
The problem is not Marxist ideology or any system of belief. It is basic human nature. Once an organization or idea reaches a certain level it becomes a monster unto itself. As Eric Hoffer said “Every great cause begins as a movement, becomes a business, and eventually degenerates into a racket.”
ReplyDeleteIt is thus not shocking that the objection is so transparently stupid. The AI can tell with 90% accuracy a person's self-reported race? So can the clinician looking at them.
ReplyDeleteYet when I consider that this means that the underlying thought is "there is knowledge you should not be allowed to look for," then it begins to make sense.
I remember watching a documentary on Machu Picchu nearly a decade ago. When Human remains were found there, those of many of the adults were quite small. This led archaeologists to believe that the last Inca king had taken refuge in his city with the Acllas Virgins who attended the sun god. But later archaeologists who studied the skeletons noted that the general body type for Quechua-speaking tribes in the Andes was different from those of people from Europe. So an average man in Peru might actually be shorter than a man from Spain. It was later deduced that the skeletons found at Machu Picchu were in fact a hotel staff who ran a royal resort.
ReplyDelete