Make a diagnosis? Elementary, Watson!
IBM Watson supercomputer from "expert" is being retrained into a medical diagnosticianDailyTechInfo
When the supercomputer, more precisely the Watson artificial intelligence system of IBM, took part in the Jeopardy contest!
(in Russia, this is called "Own Game" – VM), its massive data banks were filled with encyclopedias, works of fiction, movie scripts and history books. Now Watson's data banks are filled with medical journals, reference literature, and this has already begun to bring the first results. The supercomputer, however, with some spelling and semantic errors, independently published answers to questions in Yahoo Answers concerning questions about a strange rash on the human body and about abdominal pain. Thus, the Watson system turns from an analogue of a "connoisseur" into a medical diagnostician.
The retraining of the supercomputer has already been known for some time, but only recently IBM announced the first, still modest, achievements of Watson in this area. Watson's retraining is not the first attempt by people to create an automatic diagnostic system, such systems as Artemis and Isabel have been known for a long time. But, the incredible computing power of a supercomputer, the vastness and depth of its knowledge, the ability to understand and use natural language place Watson in a completely different league.
It turns out that the diagnosis of diseases is not much different from the process of finding answers to the questions of the Jeopardy quiz! Watson tries to squeeze as much useful information out of the question by eliminating incorrect answers from the list in advance, the supercomputer generates a list of answers sorted by probability. During the Jeopardy contest! the supercomputer could give only one answer, but in its medical application Watson gives a list of answers, along with the probability of their correctness. On Jeopardy! the supercomputer considered the answer correct if the probability of its fidelity was more than 80 percent, in the case of medicine, it is considered that there is a sufficiently high probability that the patient is affected by a not the most obvious type of disease, the symptoms of which manifest themselves, therefore Watson gives results in the list of answers and with a much lower percentage of probability.
The ability to understand natural language gives Watson the opportunity to operate with a completely new category of information for him – unconfirmed data. Such data are not absolutely true, but in some cases they are extremely useful for making a correct diagnosis. The supercomputer is able to independently "surf" the Internet, bit by bit searching for medical data with which it replenishes its data bank.
Of course, Watson is designed in such a way that it is unlikely to be able to replace a doctor with both experience and developed intuition, let's recall the notorious Dr. House. It looks more like a voluminous and smart futuristic medical reference book. Nowadays, the world is filled with information extremely quickly, any person is simply not able to assimilate something, but even find all the new information on any topic. And Watson can track the appearance of new drugs, research, journal publications, unconfirmed information and immediately use it all in making diagnoses.
Diagnostic systems based on Watson's principles are still far from large-scale use, but not far off. IBM is working on ways through which Watson's technology will be able to afford to use even medical institutions whose modest budget will not allow them to install their own supercomputer. And as more and more medical data is being digitized, Watson, thanks to this, will become "smarter" and more useful for people.
Portal "Eternal youth" http://vechnayamolodost.ru01.06.2011