Apparently respectable programmers in Japan claim to have developed an algorithm which can tell, merely by listening to a person briefly, whether or not they are about to die.
The revolutionary doomsniffer software was developed by Kenji Ohshige and a crack team of boffins at the Yokohama City University. It was developed not, as one might imagine, as part of a complicated and inadequately explained horror movie, nor as an attempt to automate the operations of so called "grim reaper" pets in old folks' homes, but rather to assist in answering 999 emergency calls from members of the public.
Apparently the algorithm was developed by trawling through six months' worth of calls and records from the Yokohama ambulance service, correlating such factors from the calls as "breathing status" and "consciousness level" with the eventual result of the call - for instance, was the caller then found dead or dying by responding paramedics, or were they comparatively frisky.
From the data set of 73,992 emergency ambulance phone-ins, almost 90 per cent seemed serious enough to the operators that triage procedures - designed to identify serious cases - were followed. However, there were only 2,349 cases where people actually died or required resuscitation on scene, showing that picking out the serious cases is difficult.
Hence the new algorithm from Ohshige and his university team, meant to help the operators tell which callers were actually in their final few minutes of life and which would last a reasonable amount of time.
Does it work? Well, according to the team's paper, published online here, the doomware has achieved "a moderate level of accuracy". ®