The medical profession is a beautiful thing. Medical professionals on the other hand are as fallible, nasty and downright pathetic as any other human being can be, and don’t let them pretend otherwise. On a related note, a machine learning algorithm developed by researchers at Stanford University has absolutely thrashed epidemiologists when it comes to diagnosing cancer.
Time constraints, communication mishaps, lack of any actual capacity to care and the fact that most of them are just festering piles of garbage, doctors tend to fail to diagnose illnesses all the time. This new AI, on the other hand, is capable of looking at slides and identifying tens of thousands of possible cancer identifiers, compared to the few hundred or so a human being can manage.
Not only that, but human doctors disagree with each other all the damn time (it’ll be that god-complex they all have) leading to misdiagnoses across the board. When left alone to analyse images, without any bias or input from the researchers, the AI was able to identify several previous unknown visual characteristics of various types of cancer. Meaning that this machine was able to pin-point warning signs that every human doctor on the planet up to this point has never even noticed.
So why is it that doctors are so terrible at their own jobs? Well, there’s the aforementioned admin and communication issues that come with a busy day, and your basic human error. There’s also the egos, the prejudice against their own patients (did you know doctors are absolutely terrible at diagnosing cancer in black patients? And that’s just one drop in an ocean of ways in which they routine prove to be useless) and the petty little foibles of man that keep them from bettering themselves at their own jobs. In a nutshell, doctors can’t spot cancer because their heads are generally stuffed too far up their rear ends.
But thanks to AIs like the one developed at Stanford University, this may not be a problem much longer. One of the great things about machine learning is that it’s capable of, well, learning. The skillset of these AIs may be specialized right now, but it won’t take much more in the way of advancement before they can learn to do literally anything a doctor can do, perfectly, in a fraction of the time. Factor in advancements in machines themselves performing human treatments, and entirely automated admin systems, and we could well be headed for a glorious future where the wretched piles of human feculence known as doctors need never poison you with their rotten coffee breath ever again.