Editorial: Are computers going to take over: implications of machine learning and computational psychiatry for trainees and practising clinicians
Argyris Stringaris
Extract
Are computers going to take over? In some ways, they already have and they are bound to take over more. How ready are we for them? And by computers, I refer to the whole range of intelligent devices that are bound to change our lives in the coming years. What do we need to do so that we can handle, trust and evaluate how computers are doing?
The first is that computers can outperform us, but we may not always understand why and how they achieve this. Back in the day, whenever I had to read a chest X‐ray, I was diving head‐first trying to find the lung tumour or the lobar pneumonia or whatever important pathology I suspected my patient had. It was the wrong thing to do. Instead, we were taught first to assess image quality, the contour of the heart, and then look at the trachea and the carina, anything but the lungs first… The right way was the structured way. It is the same ‘boring’ way in which you should look at an electrocardiogram (i.e. don’t go straight for those ST elevations in the chest leads, check the rhythm first…). I had thought then that a computer would be much better than my excitement‐seeking brain in performing such structured tasks. I was not alone in thinking that—there was a lot of research trying to teach computers to do the job of a human. It was a form of supervised learning, but it did not perform particularly well in image recognition. It is surprisingly hard to convey algorithmically simple facts such as ‘the trachea is in the middle or slightly to the right’. It requires a very long time to inject such expert knowledge into a machine and, what is worse, computers taught this way find it hard to generalize knowledge to other situations. At that time, experts could still pride themselves on doing better, and whist I was no expert, I still outperformed computers, at least with the help of the little cheat sheet I was carrying around.
We hope you enjoy the full editorial of this Issue, which is free on the Wiley Online Library.
Dr Stringaris, is a clinician scientist, Chief of the Mood Brain and Development Unit at the National Institute of Mental Health in Bethesda, MD. He trained in Child Psychiatry at the Maudsley Hospital in London (MRCPsych), and in neuroscience at the Institute of Psychiatry at King’s College London (PhD) and the National Institutes of Mental Health (Clinical Research Fellow) in the USA. Dr Stringaris’ aim is to understand why some young people become more depressed than others and how to improve our understanding and treatment of depression. He is a former joint editor of JCPP.
