Did you watch the Panorama programme yesterday on BBC (only available in the UK, I understand)? Subtitled “Could a Robot do my Job?” reporter Rohan Silva was looking at the impact of Artificial Intelligence (AI) on the workplace and jobs, primarily in the UK.
The last section of the programme was on a data analysis system at a Boston hospital (Beth Israel Deaconess Memorial Center). The reporter mentioned they use an “artifical intelligence supercomputer” (!) in their emergency department that can “forecast if you’ll die in the next 30 days”. Well, not quite, but, “forecast the probability of a patient dying with almost 96% confidence” according to the very enthusiastic doctor (and the only one featured in the programme) at the hospital. Not sure if that is all PR or verified independently.
I was very impressed when it was mentioned that the computer had 30 years of data from over 250,000 patients,so it could recognise rare deceases quicker than a doctor. After all my navigator can find me a route a 100 times faster than I can, so why not.
But then I got thinking. 30 years ago they didn’t collect patient’s blood oxygen level and blood pressure every 3 minutes like they are doing now. This was an emergency department, not the obvious place for lots of people with a rare diseases to turn up. How many rare diseases had this system diagnosed so far? So there was a fair bit of mirrors and smoke to make it look far better than it really is I think. In fact, I think the Boston system is actually just good example of what is called Big Data at work.
This tendency to exaggerate was true of the rest of the programme too which can be fairly described as sensational rather than educational.
No doubt the publicity will help the hospital. I see that the story about the dying prediction appears on many newspaper websites right now with headlines like “the supercomputer that can predict when you’ll die”!
Thanks Donna for telling me about the programme.