Artificial Intelligence in Medicine - are we there yet?

Artificial intellgence has often been viewed with a leary eye. Most experts are concerned about machine super intelligence. Some people think that full artificial intelligence could spell the end of the human race. Will we be the gods that rule the computers, the pets that are controlled by computers, or the pests that need to be exterminated? An interesting thought, but are we even close yet? Dr. Guido Germano from Los Angeles addressed the state-of-the-art in artificial intelligence in medicine at EULAR 2015. This gentleman is a cardiologist who did IT work in a previous life.

Can computers think? One way to measure this is the "Turing Test" initially proposed by Alan Turing (1950). Alan Turing is now a well-known name after the recent movie, The Imitation Game. In the Turing test a panel of judges poses questions by computer terminal to a pair of unseen correspondents, one a human and the other a computer program, and attempts to discern which is which. After 5 minutes of unrestricted conversation, at least 30% of judges must be fooled into thinking the computer is human. To date, no computer program has succeeded. The problem is that computers have trouble understanding context: "0UR M1NDS C4N DO 4M4Z1NG 7H1NG5." We can all understand this sentence but a computer could not!

Computers are really good at math, speed and reproducibility, the ability to follow clearly expressed rules, and the ability to create rules based on examples.

Let's consider the ability to create rules based on examples. If we feed the computer with data (i.e. patient demographics and symptoms) and then tell the computer the outcome, the computer will start to create an algorithm to predict an outcome based on the input. To do this we need to repeatedly train the computer with pre-determined inputs and outputs. We do this already with our speech recognition software: we train it based on our input and a correction of the output.

The issue is that these algorithm-based approaches lack context, which is vitally important. Computers can be used for a part of clinical care but we are not at the point where computers could surpass humans in overall clinical care. Think about the use of computer algorithms to help drive treatment to target in RA. Remember my favorite trial - CAMERA?

In conclusion, we are not there yet in terms of AI in rheumatology. Don't worry, we won't have Dr. HAL managing our patients!

I'm sorry Dave.

More reports from EULAR 2015