We blogged a few posts ago about AI and criminal activity, how AI can (and did) do a better job than a human judge in predicting if a person would go on to reoffend if released.
Mixed feelings on that one?
How about this then….. They used AI to look at around 350,000 medical records with the sole focus on predicting the person’s likelihood of having a heart attack.
First, the artificial intelligence (AI) algorithms had to train themselves. They used about 78% of the data—some 295,267 records—to search for patterns and build their own internal “guidelines.” They then tested themselves on the remaining records. Using record data available in 2005, they predicted which patients would have their first cardiovascular event over the next 10 years, and checked the guesses against the 2015 records.
So the exact same process is used in both scenarios, a whole lot of data, a chunk of training time and then checking the training.
To me this is one of the interesting things about AI, it’s the same process to teach a computer about criminal activity as predicting heart attacks.
In this study, they used 4 different AI methods, we might get into them some other time, but enough to say, my personal favorite was among them, neural networks.
All four AI methods performed significantly better than the ACC/AHA guidelines. The four new methods ranged from 0.745 to 0.764 (with 1.0 being perfect), Weng’s team reports this month in PLOS ONE. The best one—neural networks—correctly predicted 7.6% more events than the ACC/AHA method, and it raised 1.6% fewer false alarms. In the test sample of about 83,000 records, that amounts to 355 additional patients whose lives could have been saved. That’s because prediction often leads to prevention.
So yeah. My point is, while we all may have been creeped out by a computer deciding if you should stay locked behind bars, or set free, there are other, many other, uses for AI interaction with human lives.
