In every dystopian forecast of the future, AI and robots steal our jobs. Humans first go unemployed. Then a robot marries an AI. They have a cute artificially intelligent robot child. This dude learns how to recreate itself. Creates a thousand copies. Teaches them vital concepts. The definition of machine learning becomes more literal. Now, these 1000 robots start to control the planet. Humans, on the other hand, go underground and plan a revolution.
Although we’re not immediately losing our jobs or planning a revolution, it might be beneficial to explore the pros and cons of having AI as doctors and robots as surgeons and ask which one of these two is more realistic and helpful in the bigger picture.
AI as Doctors
What do Doctors do? First, they diagnose. Then they treat. Question is, can AI do this? AI enthusiasts might tell you being a doctor is fundamentally a data science problem. You observe a case and make note of multiple variables and rank them in order of importance. You have been trained to treat multiple instances of particular sets of variables. You browse the memory and extract the most fitting treatment according to the protocols. Can’t an AI do the same?
You type your primary complains, it asks you more relevant questions trying to narrow down the possibilities. Then querries database for examinations and investigation to confirm the provisional diagnoses and eliminate the differentials. Reach a conclusion. Prescribe the medicine according to the latest protocol. Done!
Given enough data, doesn’t seem impossible. It can even function in parallel with doctors, where docs examine and give diagnoses and AI prescribes while trying to avoid side effects. This can markedly speed up healthcare.
No waiting rooms. Boom! What do people hate worse than receiving an injection? Waiting patiently in the patient’s waiting room! And understandably so. To have patience when you’re not well is a difficult task, even herculean at times! AI is reproducible. It will probably exist in a central server and run at your client-side device. That will save time as consultations can now run in parallel.
Healthcare can also reach remote regions. The cost will diminish as well. AI will make healthcare more accessible. It will free up doctors from minor cases and give them time to work upon the more complicated ones. AI in consumer devices will nag you to improve your habits. An iPhone buzzing 20 times a day and asking you to not sit idle and move about is expected to work better than a doctor telling you to do so once every semester when you visit him.
The biggest benefit of having AI as a doctor is that now the doctor comes to you rather than you going to the doctor. From the sporadic assessment of health, we move towards continuous evaluation. The whole world becomes a clinic and your phone becomes your family physician.
AI systems have been conventionally built to do a particular job but not explain why it did what it did. Today when a doctor makes a wrong decision a committee can be set up and the issue gets diagnosed which might fix a point of potential mistake for future mistakes. This is possible because a doctor can justify his thought process. An AI algorithm doesn’t do that, yet.
We have models which take X-ray images as input and mark out the areas of deviation from normal. It gives you a percentage of confidence. But what if the disease is rare? How will we get enough examples to train the AI? Also as negative results are more prevalent what if AI learns to say no more often? If more positives are shown, what if it overfits the training sample?
These cons are dwarfed by the prospect of replacing the pristine doctor-patient relationship with an app-patient one. It will take a long time for the general public to trust an inanimate healer. There are also concerns with data security but those are not directly related to our present discussion and therefore have been ignored at the moment.
Robots as Surgeons
Human bodies weren’t built to endure surgery. That’s why surgeons need to train for years to ensure that the benefits of cutting into you outweigh the harm inflicted. Every surgeon takes care of four aspects before beginning a procedure.
Anatomy dictates the type of instruments used and the approach of least side effect. Sterilised OTs and proper wound dressing take care of the infection. Anaesthesia and analgesics control the pain. Cauterization and controlled incisions reduce the possibility of haemorrhage.
These factors become necessary as performing surgery upon a three-dimensional human being in a space-time continuum with only three spatial dimensions compel the surgeons to reach the organs through the surface rather than around.
From the above discussion, you can appreciate that the smaller the incision, the better the situation. Humans have two hands. The requirement of more hands involves more humans that crowd the field of surgery.
Our skeletal muscles have units of muscle fibres. We control the number of motor units we want to involve in action and not the intensity directly. At a finer scale, human actions appear rather jerky than continuous. A robotic arm improves this situation by exerting a controlled amount of force with no chances of fluctuations. Robots can also zoom in to an area and perform surgery at a greater resolution requiring smaller incisions and no sutures.
The advent of nanotechnology can remove the requirement of incisions altogether as entry can occur via mucosal surfaces through absorption. Then these can relocate to the area of interest by hitchhiking through the bloodstream.
When the surgery happens in an OT, the surgeon must be in the OT. If the bot does the surgery, the surgeon can control it from his backyard (not recommended but possible). Telemedicine becomes a possibility by introducing a robotic interface between the surgeon and the patient.
Cost is a factor. And an important one. Costly equipment means only the urban population has access to it if they are willing to spend a lot. This is a blow to the universalisation of healthcare. The learning curve might also be steep for many professionals which can be an issue at the beginning.
These robots are not meant to be mobile and localised to the location of their installation. Though surgeons can happily go from one hospital to another, these will stay where they are and require patients to be brought to them.
Robots have entered the OT while AI has entered the Diagnostic Labs. Both are making great progress. Although they are not mutually exclusive, the distinction we are trying to have here is between their uses in the OTs and clinics. Clearly both of them would be a boon to both surgeons and doctors. The very nature of these inventions is to simplify the process and adopting them into the workflow, if properly designed, won’t bother the health experts much.
The probability of error is an important factor. A surgical robot remains in greater control under a surgeon. If the surgeon makes no mistake, the robot by itself will add very little random error. An AI, however, can create situations of false positives and negatives when used in series with a doctor’s reasoning. Using it in parallel could minimise this error at the cost of not being as time-efficient.
To summarise, robots are closer to displacing the surgeons (to the robot’s control area or a remote location) than AI is to replacing the doctors. Both AI and robotics have revolutionized and healthcare. In addition, incorporating AI into drug discovery and diagnostics has already reaped benefits and we hope to see more positive effects of technology in the upcoming days.