Skip to content



It is often said that life imitates art. In that case, I couldn’t find a stronger case for that assertion other than the fact that IBM chose to name its state-of-the-art Artificial Intelligence Watson, an obvious reference to the sidekick of none other than Sherlock Holmes himself. In a surprising twist, that character too was named after another doctor, the eponymous Dr. James Watson. In fiction, Dr. Watson (MD) was depicted as an expert medical practitioner, but nevertheless was but a foil next to his best friend Sherlock, a literary device no doubt intended to accentuate the audience’s awe and admiration of the brilliant feats of deductive reasoning Sherlock performed on a regular basis.

However, the Watson created by the engineers and scientists at IBM proved to belie his name, for it displayed a superhuman and often astounding series of abilities to the world when it won the first prize in the game show Jeopardy, netting the machine no less than a million dollars, which it was unfortunately not programmed to spend!

In 2011, it was perhaps forgivable for a layman to think this a fluke, and not an example of the consistent march of progress that has been steadily bringing silicon chips and software within ever closer to parity with what has been since time immemorial the Gold Standard in intelligence and intellectual capability: The human brain.

And further proving the point, IBM set its creation back onto the road its namesakes once followed, the field of medicine itself. After several proof-of-concept tests, the Memorial Sloan Kettering Cancer Center in New York entered into a multi-million dollar contract with IBM to have Watson aid in performing utilization management for lung cancer cases, later stating that 90% of the nurses who have used Watson ended up consistently heeding its advice. In India, Manipal Hospitals has also partnered with IBM in the launch of IBM Watson for Oncology, a service that helps in providing personalized care to cancer patients, whilst also aiding doctors by providing accurate information even in rare and hard-to-diagnose cancers.

Now, in 2018, Artificial Intelligence, aided by its implementation through Machine Learning, is poised to take the medical world by storm, and it is in the interest of both doctor and patient to learn of its role, and how it is already, quietly but surely, revolutionizing healthcare.

A proper technical overview of all the factors that made AI software a marvel is far beyond the scope of this article, but nevertheless, I shall strive to explain the broad concepts behind its implementation, and also how to prepare for and hopefully embrace the changes it will produce in the field of medicine.

First, an explanation of the terminology:

Artificial Intelligence refers simply to any intelligence displayed by a machine, that is, any constructor system that was not ‘naturally’ formed. It can be as simple as the keyboard on your phone detecting spelling errors and correcting them so you can type faster, to the chess programs that soundly thrashed Grandmasters at the turn of this millennium, the handy little tools that let you tag your friends’ photos on Facebook, and even the self-driving cars that are constantly making the news!

Machine Learning is a way of creating artificial intelligence, using statistical techniques to make intelligent decisions even without a human explicitly programming it to do every single thing the machine eventually learns to do. Given enough time, processing power, and data, ML allows computers to do amazing things, such as learning how to convert human speech to words and then understand them, or to comb through mounds of old journals and medical data to make connections and deductions that would be simply beyond the ability of any human to find. It is the difference between a student rotely doing calculations to solve a math problem, and a teacher explaining how to solve problems, with the student then using their understanding to find answers even if the questions are not exactly the same as what they practiced on.

Reinforcement Learning is the step beyond the example given before. It is akin to a bright student trying to solve problems on their own, with the teacher only telling them if their answer is right or wrong until eventually the student exceeds the teacher and starts solving problems previously thought impossible! And yes, AI can do this, such as recently, when Google unleashed an AI called AlphaGo Zero, completely self-taught, but which still soundly thrashed all humans at the ancient game called Go, and even its ancestor, AlphaGo, which had just a few months back defeated the same masters. Go is considered a game of deceptive simplicity, but it is known to be even more complicated than chess, and thus even harder to develop the superhuman mastery shown by these AI.

So now we see how the AI is taught, let us move on to what it can do.

Of all the fields of medicine where AI and Machine Learning are being applied, Radiology is the first where these techniques are showing that they can both improve and somewhat exceed the abilities of a human doctor. Recently, scientists at Stanford displayed an AI that proved better at detecting Pneumonia than certified and well-trained radiologists! This was owing due to its superhuman attention to detail, and it being able to spot minute changes in the slide that humans simply cannot notice.

But do not fear, dear reader, for robots have not put doctors out of business yet. More importantly, these tools can be used by doctors to enhance their diagnosis, and reduce the amount of tedious busywork that has long been the norm in healthcare:

A large amount of a doctor’s time is spent in monitoring their patients, and AI has begun its rise in the field of Telemedicine, capable of doing that task and freeing up doctors and nurses to attend to other patients as well as do their real duty of treating and curing the ill.

Even the recent advances in AI’s ability to diagnose it is still another tool in the belt of a good clinician. With it aiding in diagnoses and combing through medical literature for obscure information, the doctor can take over the duty of prescribing medicine, and be applying his judgment, a factor still not superseded by machines. Much stress and effort can be relieved when it also aids him in spotting an error in the treatment protocol, such as rare interactions or genetic side-effects. With this new knowledge, the doctor can focus on the care of his patients, both increasing the number he can serve as well as the quality of service, whilst cutting costs so that even more people receive the blessings of modern medicine.

Machine learning and reinforcement learning will allow epidemiologists to churn through massive amounts of data, and by doing so find correlations and trends that let us identify and stop the next great epidemic before it takes root.

Interestingly, our attempts to make machines that are as smart as we have led us to create machines that think more like us. For example, as a way of implementing machine learning involves the use of Artificial Neural Networks, these systems are deliberately designed to mimic the way our own neurons work. They use the aforementioned network to learn from data and produce outputs in much the same way that our own brains work, albeit differing from real neurons in very significant ways.

At this moment, robotic surgery is a rather misleading term, with the name conjuring up the image of a robot cutting and slicing a patient with doctors standing bemused in the background. It is not so, for the surgical robots in current use, such as the Da Vinci are merely a more sophisticated instrument for a surgeon, allowing him or her to perform steady cuts with the precision needed for microsurgery. They more exotically allow doctors a great distance away to perform surgeries as if they were right in the operating theatre, another example of telemedicine.

Even so, advances in truly autonomous surgeries are actively ongoing. Recently, scientists displayed an AI that observed incisions made by many skilled surgeons and then used that knowledge to perform its own on a model specimen, namely the tissue of a pig. It displayed even greater consistency and accuracy than the doctors it learned from, leading to hopes that in the near future, simple procedures might be left to AI, with doctors stepping in to finish the more complicated aspects, or simply use existing means of telemedicine to finish the job from afar. Once again, surgeons are liberated, and are now free to use their hard-earned skill to perform challenging tasks, and attend to their patients. So we can see how the cost of surgeries can be significantly reduced, allowing even the neediest of patients a better opportunity for treatment, while expanding the umbrella of health services all around the world.

With this information in mind, let us imagine the near future, perhaps only as little as 10 years ahead:

The setting is immediately familiar to any medical professional today, a clean, sterile operating theatre which is suddenly the site of immense activity as a patient is wheeled in. This man, a poor farmer, was once incapable of affording anything near the costs of what his accident had inflicted on him, and perhaps would have died from the accident that cost him his leg, and now leaves him in critical condition. Belonging to an isolated rural community, even if he had been brought to the district health center, it is likely that they would neither have had the resources or expertise to take care of him. But today, long before he was brought to the hospital, preparations have been made and the staff is ready to heal him. A specialist, who would once have been much too far away in a major city just to be able to cater to enough patients to earn their bread, has been reviewing the case in the comfort of their own home, as nurses and assistants run IVs and prepare anesthesia. He or she needs not rush and is able to review the case, because AI of the medical robot is already preparing the patient, cutting away necrosed tissue, whilst other software admits just the right dose of necessary medication without careful monitoring by human nurses being needed. The nurses can then rush to the IPD or Emergency to attend to other patients, knowing that their ward is being monitored and they will be immediately summoned if vital parameters are judged to be abnormal, or even if there is an expected risk of complications arising. The doctor dons a virtual reality headset and is immediately in control of the robot. They now are in possession of data that modern surgeons can only dream of, for in front of their eyes they can see hidden anatomical features, such as hidden arteries or veins, even though dirt, blood and obscuring tissue, with the AI intelligently predicting and tracking relevant details through its advanced sensors. They can take control at any point, with the software compensating for tremors or micromovements that produce small errors in the motions of the doctor’s hand, allowing for pinpoint precision, or order the robot to perform procedures carefully and correctly. Ideally, this intervention would have saved the patient’s leg, but due to the delay in his arrival, we must settle for saving his life.

Once he is in the surgical ward, he is still safely under watch, medicine personalized in both dose and type is being provided, novel antibiotics produced as a response to the modern risk of antibiotic-resistant infections are at work. In fact, in all likelihood those drugs were invented by a pharmaceutical AI, that synthesized them after doing trillions of calculations and simulations to find a way to overcome bacterial adaptations, often in advance of them occurring. They were also deployed so quickly clinically because such AI can help skip the need for protracted drug testing and clinical trials, as well as help in follow-ups with the possibility of catching subtle effects in the future.

The benefits of AI will not cease even when the man is discharged, in our day, he might have had to settle for a crutch and the resulting disability for the rest of his life. However, the same advances in modern science that ensured that even a relatively backwater hospital could afford a surgical robot as well as the access to a specialist doctor now allow him to afford a prosthetic leg. It shares but the name to the clunky, and inferior models we use today. This one can be integrated into the patient’s body to a far greater degree, a relatively cheap device reads the electrical outputs of the neurons that once lead to his leg muscles, and AI uses that knowledge to ensure that moving it feels as natural and intuitive as his real leg did. He soon walks out of the hospital a healthy man, and thanks to both the surgeon and advances in medicine for securing his safety and good health.

This tale might sound fantastic, but I assure you, dear reader, that it is not. The techniques and technologies involved are very real and are already being deployed today. As with all technology, time will only make them cheaper, and accessible to all.

So now we come to the real issue, how do we, as doctors and aspiring doctors, prepare for this future? I will state a few points that are relevant:

  • Healthcare will become much cheaper, this will allow a much larger number of patients to be inducted into the systems already in place. Even though we might fear these AI taking our jobs, it must be admitted they are also increasing our total number of patients, and thus the need for doctors again!
  • Right now, individual clinical skill and bedside manner share equal importance for a doctor’s success. This might sound surprising, so I will refer you to the studies that discovered that how often a doctor gets sued for medical malpractice does not even depend on their skill and patient outcomes, but actually on their rapport with the patient! Even a below-average practitioner ultimately provides the most patient satisfaction by taking the time to get to know their patient, and by assuaging their concerns. Patient satisfaction is a concern that is sadly ignored much too often, to the detriment of many a doctor.
  • AI assistance will greatly decrease the skill gap between the best doctor in a given field, and the baseline. Right now, a doctor’s knowledge and experience can help him significantly exceed the performance of his peers. However, with AI catching the hundreds of tiny issues and esoteric bits of knowledge that an experienced clinician currently knows, that gap will be significantly reduced if not eliminated.
  • Thus, a young doctor today would be wise to pay more attention to their soft skills! It soon will become the most visible aspect of their practice, at least to the patient and the public.
  • Today, the driving pressure on doctors is to specialize in ever more niche fields, but AI will largely remove the need for such hyper-specialization. There is likely to be a massive rise in the demand for General Practitioners, but the rate of this change is sufficiently unclear that students should not just abandon their current MD courses!

Thus, we come to an end of this overview. I hope that I have convinced you that the future holds just as great an opportunity as it does risks and that these guidelines will prepare you for it. This has long since ceased to have been a matter of if but rather when. The mark of a great doctor is their ability to deal with rapid change and information overload, and our training will definitely be needed to weather the road ahead!

Leave a Reply

Your email address will not be published.