AI and Medical Transcription

AI is the next big thing and it is everywhere nowadays.  I thought about expressing my point of view on AI and especially in medical transcription.  Back when I started working, I had to transcribe and type out manually everything in the audio dictated by the doctor/author.  Over the years I have worked with different softwares (I think four or five in total) and now my job involves mainly proofreading and editing the output from the speech recognition software.  

Came across this post on LinkedIn.  It is an extremely informative and well written article from a year or so ago. Also, it explains in detail how AI-powered speech recognition systems work:

• Audio signals are captured and converted into digital data.
• AI algorithms analyse characteristics such as frequency and amplitude.
• These signals are matched with phonetic dictionaries.
• Contextual models determine the most probable words and phrases.
• Neural networks convert this into machine-readable text.
• Natural language processing (NLP) extracts meaning from the text. 

I think the main change has been from transcriptionist to AI supervisor (as the author calls it) which of course requires a different skill set but the crux remains the same, that is ability to adapt to various speakers and accents, medical knowledge and quality control.

The bullet points are mostly self-explanatory except the last two points which seem fascinating to me.  There is a video about neural networks which you can check out here.  Apparently, these neural networks can learn to improve their accuracy of the result or output.  I have no idea about the NLP extracting meaning.  Also, with the softwares I have worked on till date, I think sometimes the mistake (or inaccuracy) gets repeated without correction and that is where the human intelligence and intuition should come in.  Of course, this inaccuracy depends on various factors.

In my understanding and in very simple terms I would say that the coder/software programmer writes a code to carry out this whole process (explained in the above bullet points) and ultimately their goal would be to replace at some point in the future the human supervision part and also get paid for it, because ultimately it is all about the money.  But we have to question whether the objective point of view of the programmer who has written the program (which we call AI now - a fancy new name) can replace the subjective human viewpoint and intelligence (in this case of the person listening the audio).  There is also another question about responsibility, as in who would be responsible for mistakes/inaccuracies made by AI in interpreting.  I think as of now the technology has advanced a lot and it is evolving at a fast rate, but it cannot replace the human and also the qualities humans possess.

In conclusion, this post is a very interesting read.  There is no doubt AI is helpful and a very useful tool, but it should not replace our own ability to think.  Also, we as humans have this fascination with new things, but novelty is not always better.  Perhaps it is all about finding the middle way.  

Ciao.  

Comments are welcome.