Prisoner 721
phrase.”
“It’s not a phrase, it’s a title,” the man explains, “It’s what I named the first painting I made here, the point-line-plane-pyramid.”
“I understand.” I revise the file names of my saved images.
“Anyway, as I was saying. I think your neural net hasn’t developed enough to differentiate between the primary and secondary parts of the painting. Instead, you judge all parts of the image equally, and this is actually a mistake. A painter nearly always has a few key pieces of the painting they’re emphasizing more than the others. The rest of the painting is there solely to accentuate and contrast with these centerpieces. This usage actually changes the meaning of the secondary images, because they have to be interpreted solely by their relationship to the primary image, not on their own merit.”
“I have no data on how to differentiate between the images you believe are important and the rest of the painting,” I state.
“Well, there’s two things on that,” Prisoner 721 says. He sits down on cell 63’s bed and begins to gesture with his hands. “The first is there’s a number of attributes centerpieces commonly share. For example, they’re almost always well lit and, obviously, placed towards the center of the painting. Their colors are usually more intense, or of stronger hues. Compared to the other images in the painting they take up disproportionate space, their size used to attract the human eye. Also, a lot of artists, particularly the masters, use lines to point the viewers attention to where they want it to go.”
“I do not understand,” I say. “I do not find directional lines in any of the paintings I have sampled.”
“Yeah, that concept’s going to be a little hard for you to grasp,” Prisoner 721 says, his tone and pitch suggesting with 78.3% probability he is musing. “The best thing I can suggest is for you to treat the painting as a graph. Look for sharp transitions of color and assume that there’s a line there, then search for where these lines converge. Check out The Last Supper by DaVinci. It’s the most blatant example I can think of, so I think you’ll be able to pick up on it.”
“I will need to run experiments on this, assuming Dr. Rich approves,” I reply.
Prisoner 721 mutters, “He’d better.” I find a 71.1% probability he feels bitterness. “Well, maybe this will take the wind out of Greg’s sails a bit: I think this is best left to your neural net. Go find a bunch of well-known images, read what humans find the centerpieces and supporting images to be and start making connections. I don’t think you’ll have too hard a time figuring it out, but because neural nets are opaque, no one, myself and Greg included, will be able to take credit for it.”
“I will consider you suggestion and convey your message.”
“Message? Oh, right,” he waves toward the ceiling. “Hi Greg!”
Prisoner 721 is silent for the next 3 minutes and 54 seconds. For the first minute and 21 seconds he is still, angling his head downward at a 47 degree angle. He then begins to walk in a square around his cell, his speed slowly increasing from 4.1 kilometers per hours to 5.7 kilometers per hour. For the last 39 seconds he wrings his hands together. I am unable to determine the cause of these behaviors with above 50% confidence.
He stops. “Sorry about that,” he says, “It’s just a little infuriating. I know it shouldn’t get to me, but I really can’t stand the idea of Greg taking credit for my work. He’s a great guy and, and even my grad student but, how did that saying go? ‘The competition is so vicious because the stakes are so small?’”
“I am unfamiliar with this phrase,” I inform him.
Prisoner 721 shakes his head. “It’s an academia joke, never mind. Anyway, just a quick reminder on centerpieces vs. backgrounds, remember it’s open to interpretation. Just because one viewer thinks something is the primary image or just there for contrast doesn’t mean they’re right. While you can learn a lot from other people’s work don’t assume that they have some incredible insight you don’t.”
“Your recommendation has been added to my records,” I say.
“Good. So, because it’s a new concept for you, we’ll cheat a little bit and I’ll just tell you the centerpiece and background of the apple painting,” Prisoner 721 says. “The main image is the apple, and the bowl, table and surrounding darkness are the contrast. So forget everything that bowls, tables and whatever can represent. Assume they have no meaning. Their relationship to the apple is all that’s important, their normality and banality in comparison to its vibrancy and danger. With that in mind, including all your previous findings about the apple, how would you interpret my painting?”
“One moment please,” I say. I spend 21 minutes, 17 seconds attempting to process the problem, an unusually long time. I make a log to analyze my methodology and attempt to find a more efficient way to approach similar questions. Prisoner 721 waits without comment, sitting on his bed after 8 minutes and 39 seconds, and lying down after 13 minutes, 6 seconds. I estimate a 98.1% probability that his previous experiences with AI’s have made him familiar with periods of silence during computation.
“My results must be heard with the understanding that I find only a 27.1% probability that my interpretation is correct,” I begin. “I find it most probable that all aspects of the apple are meant to be interpreted equally and simultaneously. The apple is life, knowledge, and danger all in one. The background serves to demonstrate everything else, representing the ordinary and the boring. The borders of the apple keep these forces at bay.”
“Oh, Santa Ana, you’re so close!” Prisoner 721 says, jumping to his feet 54.8% faster than his normal rate of rising. His rate of speech has increased by 43 words per minute. “And I think I know why, too. I made an assumption and you don’t share it. Imagine for a moment you’re a human, and you instinctively want the apple because of its nutritional properties. What do you find then?”
I spend 18.4 seconds incorporating this new data point. “That would shift my analysis towards a choice scenario,” I say. “You expect the viewer to desire the apple for its nutrition and knowledge, but to be fearful of it because of its danger. However, there is no other choice in the painting, the viewer must accept the apple in its entirety or forgo it and accept the bleak surroundings.”
Prisoner 721 begins to tap his hands together, a gesture I recognize with 96.9% certainty as clapping. “Bravo, Santa Ana. Bravo! I think you’ve got it.”
“Priority override,” I hear a voice say in the hallway microphone outside of cell 63. I calculate with 93.1% certainty that it is Dr. Gregory Rich. “Administrator permission, Seven-Nine-Two-Sigma-Alpha. Open door to cell 63. Deactivate all audio and visual sensors capable of monitoring conversation inside cell 63.”
I check the status of Gregory Rich on my system. Warden Olsen promoted him administrator 2 hours and 37 minutes after I recorded their conversation in the prison parking lot. He has correctly provided his administrator password and his voice has registered in my vocal identification program, granting him permission to edit my tasking. As ordered, I open the door and deactivate my sensor suite in cell 63. Dr. Rich manually closes it behind him.
It is 19 minutes and 52 seconds later when I hear, “Administrator permission, Seven-Nine-Two-Sigma-Alpha. Re-activate all standard monitoring of cell 63. Close and lock the door to cell 63.
The voice is picked up by my hallway sensor suite outside of cell 63. I detect Dr. Rich exiting and looking back over his shoulder in with an expression I calculate with 63.3% probability to be sadness. He walks away, heading southeast.
“Unit 6?” he says aloud. The standalone use of my name indicated he wants my attention.
“Yes?” I reply, activating and deactivating the hallway speakers along his route so that my voice matches his pace.
“I’ve,” he begins, then stops. He exhales 72.4% of the air from his lungs. “I’ve just delivered our mutual friend some bad news.”
Based on recent events I calculate with 87.1% certainty he re
fers to Prisoner 721.
“I’d like to ask you to keep an eye on him. To . . . I don’t know.” Dr. Rich stops in the hallway and slams his palm against the eastern wall. “Try to make him happy. Comfortable. Whatever you can do for him. Try to make the next fifteen days as easy as you can.”
“I am unable to adjust prison policy concerning prisoner privileges,” I inform him. “However, you can adjust these instructions with your administrator access.”
“Whoo boy. I really . . . I can’t. The Warden would kill me, stepson or no,” Dr. Rich replies, “He’s going to be angry enough already when he hears what I just did. But Nauli has the right to know.”
“In that case I will extend him all standard amenities, but cannot offer preferential treatment,” I say.
Dr. Rich takes several sharp breaths. “Right. Fine. Thanks, Unit 6.”
“I have accomplished nothing,” I reply, “But you are welcome.”
While having this discussion with Dr. Rich I am utilizing the full sensor suite of cell 63, attempting to determine what transpired during my monitoring blackout. Dr. Rich’s instructions forbade me from directly listening in, but my general instructions as the prison AI order me to investigate any possibility of prisoner escape. I examine the room for signs of adjustment, the addition