Current research estimates AI’s error rate of up to 23.31%—and this becomes especially dangerous when AI is used in such high-stakes industries like healthcare. Even though AI enabled the possibility for automatic speech recognition systems (ASR) to provide very quick transcription results, the speed often becomes not the most important factor. When accuracy is mandatory, human-powered transcribing still has the upper hand.
Even though AI-based transcription technology has improved tremendously since the turn of millennia, offering cheap and fast results, the error rate has to be roughly halved until an ASR system can reach near-human levels of accuracy. A study found that mistakes were made by speech recognition technology at a rate of 7.4%. They commonly were incorrectly prescribed dosage of medication, and wrongly transcribed names—an identified example of which would be mixing up Celexa, prescribed for anxiety and depression; and Celebrex, a medicament for pain. Such errors might have dangerous consequences.
“AI cannot handle difficult medical terminology which results in inaccurate transcription,” explains Mindaugas Čaplinskas, the CEO of GoTranscript, a company that transcribes around 5000 hours of medical content per year. “This is unacceptable in the field of healthcare—having someone in-house to review and edit it takes tons of time. Despite companies thinking this option is cheaper, in the end, they spend more time and resources compared to the more efficient alternative of outsourcing it.”
Medical practitioners mostly use transcriptions to assist them in their daily tasks, such as patients’ medical notes and disease history. Here AI makes the most mistakes in deciphering terminology of the medications, diseases, equipment, and terms in Latin, resulting in critical errors such as misdiagnosis. Meanwhile, a human transcriber can do research and identify the exact word used, written correctly, and outperform AI at an accuracy of up to 99.8%.
“Transcribers can fact-check the information, they adjust to different accents and dialects, and easily work with audios that contain background sounds or filter out fillers and speech errors,” clarifies Čaplinskas. “This all adds up to a more accurate transcript.”
Day by day, an increasing number of medical institutions choose to use transcription to assist them in their everyday work to save time, operational costs and increase internal productivity. However, automated yet inaccurate solutions might bring more trouble than help that might even result in misdiagnosis. That’s why, up to this day, human-powered transcription is prioritized by medical professionals.