Flowers: just for you, I'm going to do some copy typing from a .jpg of three pages of a confidential document from an AI company, showing their strategies in tracking and manipulating gamers in their own homes to carry on playing and to buy more game time.
You'll see at the end that the AI start aggressively targeting women in the 48 hours of the luteal phase of their menstrual cycle, as detected by the pitch of their voice. “Obviously we had to turn it off” they say, but this is the only time they express any concerns about tracking and manipulating gamers' behavior.
Now, what they describe as “Our Personal Favorite” example is also interesting to me. This outlines what they can infer when there are a lot of chair-scraping sounds.
Interesting, also, in that it is the only part of this confidential document that is highly redacted.
I've made a claim on FR and elsewhere, that I developed the only properly validated protocol for measuring “attention” in the classroom. As it happens, I had a big, big fight with the university over the dissertation in which I describe this instrument. It's a long story. I decided that the the thesis was too dangerous to my health, and I withdrew it from the library. It's a long story.
So I've sat on my little secret for over 30 years now. I periodically check whether anyone has developed some way of measuring “attention” in the classroom, but I can't find any references to such a validated instrument.
This is interesting, because some notion of “attention” is vital if you want to understand what's really going on in a lesson. You may record the teacher saying something, and transcribe it, and say: “This is what the teacher said.” But was anyone LISTENING? How do you know?
The only instrument I found in the literature was from the Ford Project in the UK, where they took still photographs every 30 seconds, and counted the number of kids who were looking up at the teacher.
This is next to useless. The kid who's writing something down may be paying far more attention than the one gazing blankly ahead. There was no attempt to validate this instrument.
I worked solely from audio recordings, with written notes from what I'd observed from the back of the classroom. I had a microphone at the front, with the teacher, and one at the back. It provided quite a weird stereo picture of the classroom. I transcribed every word I could hear, a really slow and tedious process, but essential to understanding what was going on.
Eventually, I became acutely aware of the background noise — feet shifting, pens tapping, rustling, and chair scraping. I became aware that at moments of clearly high focus — like when a very strict teacher had just yelled at the class to pay attention — the class would go completely silent, and this background noise would disappear.
I marked these episodes in the transcripts, and checked multiple times with the students and the teacher, who confirmed how the whole class would focus on the teacher's exact words during these moments — “You have to listen to every word he says, you've got to get it exactly right, or he goes totally mad”, was one comment.
So in the end, across a variety of classrooms, I could calibrate the level of attention directly from the tape recorder's sound level in decibels during pauses in the teacher's delivery. If there was lots of shuffling, scraping and general restlessness, attention was low. I can go back to those classrooms decades later, just from a recording, set a baseline, and put a number to the attention level at any point in the lesson.
I promise you, I can tell the difference from these recordings between a kid staring straight ahead at the teacher who's pretending to pay attention, and a kid looking up who is really listening. No matter how much you try, you'll start fidgeting a little if you're bored or distracted.
So, that's my famous protocol. And now, finally, I discover that an AI has rediscovered it. This is why I'm putting it forward publicly here for the first time, I want to claim it before the machines do.
What really piques my interest: why is this the only redacted part of this leaked document? What big secrets did they find, correlating with chair scraping?
And Flowers, please note: this is just what the AI derives from the SOUNDS inside your room. The rest of the document is about how it tracks your devices as you move around your lounge, using electromagnetic imaging. This is under the heading “Mapping Users Homes”.
So when I say: these systems monitor you, to an unbelievable degree, do you believe me now? This is just a games platform, finding ways to track and manipulate gamers and feed their addictive behaviors to make money. And it's tracking you inch by inch as you move around your own lounge.
This isn't the military surveillance, this is just the gamer companies monitoring you.
Put it all together, hundreds of companies monitoring you, multiply this by billions of people, and remember that all this vast stream of information is being recorded, so that patterns in your behavior can be identified and tracked.
We will be swimming in oceans of our own data, streams of information that are constantly used to trick us and game us and manipulate us and identify our weaknesses and exploit them. This is (I hate to say it, but it seems to resonate with people): The Matrix, for real.
OK, enough intro. This is the first time this document is being made available as a text version, as far as I can tell. I cannot tell you where it's from, what company is involved, there were no details, this was just dumped on the QResearch board, you find the weirdest shit there. But this cannot be a hoax, it's just too detailed.
This is how the algorithms are really gaming the human race:
—>Schedule “Z” Audio
Raw information – Detail: Derived Information
Audio – Processed keyword analysis and non-word sounds: The library of sounds we've crafted was made by us gathering sounds from other companies and creating many ourselves. The list of identifiable non-word sounds is currently over 500,000.
This includes sounds we consider significant to monitoring the user (door slamming, yelling (pain, anger, sadness), laughing, falling down, children crying, television show dialogue) and insignificant sounds (rain, rustling from readjusting sitting position, eating (microwave sounds followed by chewing sounds indicates eating reheated of prepacked food. Also includes guessed food types like potato chips, soup, spaghetti)).
The library also performs mathematical transformations to detect synchronous vibrations or sound reflective surfaces to determine the size of the room the user is in. It can sometimes even detect if there are open windows despite no sounds coming in through them.
We are also in the process of contacting the author of
http://jmlr.csail.mit.edu/papers/volume14/stowell13a//stowell13a.pdf which with their help in applying their algortithm [sic] listed in the paper will help us to pinpoint certain variables we require.
Example highlight: In one case, with just a large dogs [sic] footfalls and barks we were able to guess when to stop a game excitement time because the dog needed to go on a walk outside of the normal walking time pattern.
The user approached their home again, when they approached a sedentary position (guess: couch) then the AI sent a notification of a large bonus to the user to encourage a new game session. The user began a new play session.
Example Highlight (And Our Personal Favorite): The AI found a correlation between chair scraping noises and high quality, low quantity spenders. If someone has a lot of chair scraping noises, they likely (HEAVILY REDACTED)'
Using voice tone and pitch, after determining face and gender, we can detect more than just moods. It can tell why the user is feeling that way. The AI started detecting, and then aggressively targeting women during the last 2 days of the luteal phase of their menstrual cycle. It discovered a correlation between the voice pitch adjustment away from the normal standard deviation, and that women would buy more in those 48 hours when the AI used aggressive upselling/advertisement strategies.
Obviously this had to be turned off, but when we were in the process of confirming the recognized pattern we were surprised. At the time it was given the goal with a high point score for causing ad fatigue and harsh ad experiences, harsh ad content, and intrusive ad placements at time of detected frustration in game.The specifics [sic] audio queues it derived were similar to
http://journals.plos.org/plosone/article?=10.1371/journal.pone.0183462. [Link does not seem to work.]
THE INFORMATION CONTAINED WITHIN THIS PAGE IS CONFIDENTIAL. <—
