Unlike most sciences, medical science needs experience of treating patients. Patients diagnosis and response to treatment depends on a uncountable number of variables, some are known and most are unknown. This leads to a lot of uncertainty or probabilities while managing patients. So, medical field depends on statistics for diagnosing and managing patients.
Analysing patients symptoms and signs need a kind of intuitions for doctors because patients tend to describe or develop a variety of signs and symptoms for the same disease. This symptoms and signs, of course, follow a pattern but many times with a lot of variations.
In recent times there is a big buzz around machine learning and artificial intelligence. Although I am not an expert in the AI or machine learning, I have developed a kind of attraction towards and seriously thinking to learn this machine learning.
Most classical software programs obey what has been written in, they hardly have the capacity to deviate away from algorithms, this leads to a kind of limitations for software developers when dealing with a very large number of possibilities and very big kind of algorithms. In such situations, intuition luck and previously gained experience become advantageous like we human do.
In the medical field, many a time for the same patient with same disease different diagnosis and treatment is advised by different doctors. The same doctor may suggest different treatment for the same patient when shown at different times. This happens partly because medical field deals with a very large number of possibilities and us humans do not scan all possibilities but pick few while dealing in such situations.
Although Software programs have the immense capacity to iterate big algorithms (but there is a limit for this also) but to program them to do so in the traditional way is very difficult and hilarious. The easiest and probably the best way is self-learning by exposure as we humans do.
Google deep mind AlphaGo.
‘Go’ is a game played in china, similar to Chess. Unlike chess in Go possibilities of next move are more than a number of atoms in our universe. So the computer can not compute them in the traditional way. Human Go player depends on intuitions. AlaphoGo a program from Google has beaten south Korean world champion Lee Sedol in march 2016. link.
The IBM chess computer Deep Blue, which famously beat grandmaster Garry Kasparov in 1997, was explicitly programmed to win at the game. But AlphaGo was not preprogrammed to play Go: rather, it learned using a general-purpose algorithm that allowed it to interpret the game’s patterns. link . This is similar to human behaviour.
The implication of AlphaGo on medical science.
We have already seen better AI in diagnosing cancer as compared to the human pathologist. If we create programs which can learn the surrounding by observing and experiencing events without a traditional input of commands then it will make a huge difference. This self-learning capability now does not appear like any science fiction, it is already a reality.
It is predicted that in next thirty to sixty years AI will become part of our life. I personally feel Robo-DOC like the one we see Hollywood movies will be a reality. Sometimes I feel they may excel humans because for them to transmit already learnt knowledge is a very simple, i.e. push of a button, unlike we humans, have to undergo intense training.
For computers, size is no limit, unlike we humans, have cranium as a limit, and to make a network of computers is very easy this can increase processing and size also.
Human to human transmission of knowledge is a hilarious one. We can not be plugged and played, unlike computers. What one computer has learnt by exposure to an environment can be easily transmitted to millions of computers, intern millions to millions just by connecting.