Nearly 50 years after the iconic film, 2001: A Space Odyssey, was released, the spine-chilling words, “I’m sorry Dave, I can’t do that.” live on. Spoken by the infamous, anthropomorphized computer, HAL 2000, the characterization—and potential dangers—of a seemingly flawed artificially intelligent machine, even today, remain an active topic of debate amongst the greatest, living minds on the 21st century.
Artificial Intelligence aside (sorry HAL), traditional statistics and more recent paradigms such as machine learning are commercially used, today, for everything from credit card fraud detection to search engines such as Google. In fact, while statistics and machine learning were born out of different sciences and use vastly different vocabularies, they are simply two important tools in the greater field of information science. According to statistician, Larry Wasserman, they are both concerned with how we learn from data.
And therein lies the problem and opportunity… data.
It is not that we record too much information, it is we record more than we are able to effectively consume. In fact, significantly more information has been stored since the turn of this century than had been previously recorded throughout the entirety of recorded human history.
So why all the hype around machine learning and medicine? Simple, it is an extremely fast and efficient way to learn from data—lots of data. And it is precisely the fact that medical professionals so dutifully record information regarding patients—from observations to test results—that makes machine learning particularly useful in the field of medicine.
Machine learning, therefore, does not look to replace medical professionals, it’s poised to assist them.
The recent shift in attention towards machine learning is driven by a massive increase in the scale and breadth of data we are collecting and preserving. While the processing and storage capabilities of modern computers continue to grow, even the bleeding edge of technology, for some of our newest, massive data libraries is approaching intractability.
Machine learning represents one relatively new tool in our data mining arsenal.
Despite all the excitement, doctors and medical professionals should not be surprised to know that these new learning machines are not simply assistants with the ability to read thousands of cases a minute. They are, in fact, just as likely to help better serve patients through scheduling and the prevention of hospital re-admittance than help cure the common cold.
Diagnosis and detection are clearly one facet of medicine currently being tackled by machine learning such as IBM’s Watson, Google’s DeepMind and Microsoft’s Cortana projects. But insights gained from machine learning are likely to impact doctors and patients in other tangible ways as well.
SMARTMD, for example, is working with machine learning to even further improve transcription accuracy and streamline the instantaneous filing and recall of patient records. Much work behind the scenes, our goals are to reduce the time and cost while improving the outcome and experience of every patient visit.
While it has a huge potential to contribute to the practice of medicine, machine learning, in and of itself, is not a tool for innovation. Neither, for that matter, is statistics. For the time being, innovation remains one of the great gifts of human consciousness and, just perhaps, the distant future of artificial intelligence.SHARE