Part of my job is trying to predict the future (it’s an awesome job). Occasionally I read something not necessarily relevant to investing but rather to some facet of life.
With that in mind, my thoughts turn to computers that aid in medical diagnoses. Consider for example an EKG (aka electrocardiogram, which is a tracing of a heart’s electrical rhythm). Computer algorithms can interpret these tracings to determine such things as whether a heart attack is taking place or has occurred in the past. The computers are only about 80% accurate (results should definitely be reviewed by a cardiologist), but that’s actually not bad.
As we (the human species) make technological progress, we’re developing increasingly complex algorithms, ones that mimic human learning. One of the first examples I saw of this was Google’s translate function (www.translate.google.com). If you’ve never used it, it’s worth checking out. What is fascinating is that the program does not simply translate words one at a time. That methodology (translating word-by-word) can lose some of the ‘art’ of language (i.e. trying to translate idioms such as “it’s raining cats and dogs”). Instead, Google looks for patterns of word associations across millions of documents. In doing so, Google is not translating the text but rather translating the ‘meaning.’
Recently, there was a wonderful article in the NY Times discussing computers that are designed to learn.
Google researchers were able to get a machine-learning algorithm, known as a neural network, to perform an identification task without supervision. The network scanned a database of 10 million images, and in doing so trained itself to recognize cats.
If computers can recognize cats in an image, why not other things? Perhaps because my focus is medicine, the immediate application that comes to my mind is reading X-rays, CT scans, and the like.
For those who have just entered a radiology residency, there’s no need to hit the panic button. Just as with computer-assisted EKG interpretation, we’re almost certain to want a human radiologist double checking the machine’s work. The computer is likely to make some mistakes, and there are also likely to be medical problems that a radiologist can spot that a computer was never trained to look for in the first place.Suffice to say that I don’t really think radiology as a field will be threatened (it may even come to pass that computer aided detection will improve how radiologists enjoy their work). Even so, perhaps before choosing radiology as a profession, it is worth at least taking a moment to contemplate the future.
For now, the technology is nowhere close to what would be needed. The NY Times article describes that ‘learning computers’ are still pretty limited in their abilities:
I.B.M. announced last year that it had built a supercomputer simulation of the brain that encompassed roughly 10 billion neurons — more than 10 percent of a human brain. It ran about 1,500 times more slowly than an actual brain. Further, it required several megawatts of power, compared with just 20 watts of power used by the biological brain.
Perhaps another way to read this is as a testament to how incredibly amazing our brains are.
Before we become too self-congratulatory about our intelligence and the processing power of our human brains, I’ll close with a quote by Albert Einstein just to keep us humble. “Only two things are infinite: the universe and human stupidity . . . and I’m not yet completely sure about the universe.”