HALLOWEEN. It's right up there with Easter and April Fools'Day in the pantheon of crap holidays, and it's back to irritate us and make us go and hide in a dark room till it's over. Which is sort of mission accomplished, ironically.
Anyway, in an attempt to enter into the spirit (ho ho) of the haunted holiday, take a look at this experiment to see if a Raspberry Pi can detect ghosts in walls. Which of course it can't, morons.
The "ghost hunting machine" is running live throughout Halloween and uses a combination of a Raspberry Pi, a custom-written Node JS app which feeds into the Google Speech Neural Network API and then displays the 'results' on a static background.
Because it's all very series and scientific, all the results are stored in an SQLite database for later ‘study',
The Red Pepper team explains: "The power that modern electronics (like EMF meters, thermal video, and audio recorders) give us is good, but still leave a lot open for interpretation. Since pareidolia* accounts for most "proof" of paranormal activity, we built a system to objectively listen and classify Electronic Voice Phenomenon (EVPs), making human interpolation of noise a non-factor.
"We're employing Machine Learning to prove once and for all if Ghosts are real or not**, using the latest in Speech AI backed by a Deep Neural Network on a scale never before possible."
So if you want to see if all this ‘hocum' has any credence, tune in to the live feed below, broadcasting straight out of Germantown, Nashville where absolutely nothing will happen. Unless… what's that Dolly? You… want me to pick you up? Why do you want me to do th… ThIS iS tHe VoICe of THe oTHer reALM. bLaCKbeRRy iS thE bEsT OpeRaTinG sYssTeM. YoU mUsst aLL bUY BLaCKbeRRyS… µ
*either this means ‘finding patterns where there are none', or it's a sort of gladioli, we're checking.
Own Black Friday with our list of this year's best offers
Wistron is reportedly gearing up to begin production
The move is set to happen by 2020
You'll definitely be able to get one in time for Christmas