Roseworld Transcript S1E05: Charlestown

Broadcasting in three, two…


Continuing reading 759 of the truth. Try to forget what is known. Something else is knowing it for you. There is a reason it has to be this way. Telling you outright… you’d never believe it. If you are waking up, or you’re in-between… I’m sorry. This is no world to be awake in. Find a safe place if you can. They do exist. You can clear your mind there. If you can’t find one… act natural and keep looking.


Let the reading begin.


TABPAGE ARTICLE: May 22, 1984. Central New York Facts.


If you live in or around Utica, you may be familiar with the Charlestown Mall. It might even be a favorite shopping destination of yours. This gorgeous brick-wall metaplex has been credited with bringing new life to Central New York with its state-of-the-art facilities, a wide selection of international goods and services, and its high-tech friendly terminal intelligent systems. But did you know this complex of buildings is not new, and has been in use for quite some time?


What is now known as Charlestown Mall, or Charlestown USA, began life in the 1920s as a weapons manufacturing plant for the United States military. Popular conspiracy theories hold that the factory was responsible for more than just the usual ammunition, especially during the Second Great War. Some think that the same high-tech weapons experiments that supposedly were conducted in the 1940s in rural Pennsylvania in fact got their start in Utica. These rumors only gained more traction when the plant, converted in the 1950s into an autab parts supplier, entered into a contract with the analog systems engineers in Union County, the same area of Pennsylvania which had created so much intrigue in the first place. In reality, many autab manufacturers and research facilities were contacted by Union County, which had made great strides in the field of neural network implementation. 


Machines that could truly think - though not on the level of any man. That was what the future mall helped develop. And the building’s owner, too, would eventually benefit from the technology they helped popularize. When Charles Gaetano converted the manufacturing facility into the shopping center it is today, the infrastructure was already in place to create an incredible high-tech shopping experience.


INTERNAL DATA: April 7th, 1987. Encounter recorded in Timothy Gold and Jewel storefront within Charlestown Mall, processed and compiled by the Titan Utica. 


Afternoon. Little heads watch a woman approach. They have seen her before. She feigns brief interest in the various trinkets bathed in the store’s warm lights as she approaches the counter.


“Hello,” she says. “I was sent here by another jeweler. I’m told you have the best mineral identification process in the state.”


The man behind the counter eyes her up. Little heads watch him wipe a sheen of sweat off his brow. “Sure.” He pats a machine, filled with dozens of little heads, and grins. “This’ll tell you if what you’ve got is worth anything.”


“It’ll give me an identification?” the woman asks, frowning.


“Sure. What are you looking to sell?”


“I…” the woman hesitates. She pulls a strange rock out of her pocket and reluctantly hands it to the man. “Can you tell me what this is made of?”


He turns the object around in his hand, holds it up to the light. It’s uniform gray, almost metallic if he turns it just right, but part of it is pink, like it was chipped off of a geode or some larger formation. That pink part almost feels… What does it feel like? Why does beef jerky come to mind? For some reason a chill runs through him. “The hell is this? This just a rock you picked up outside?”


“Can you tell me what it’s made of?” she repeats, almost snapping. Then she says, gentler, more politely: “I have reason to believe it’s made of something strange. Uh, valuable,” she adds, almost as an afterthought.


The man looks uneasy. He feels uneasy, too, like he should be scared of something. But he knows better. This is his store. This woman is too timid to try anything. And the thing in his hand… it’s just a rock. It’s a weird rock, but it’s just a rock. Pedestrian. He flashes her a grin, then places the rock in his analytical machine.


“It takes a few minutes,” he says, then leans on the counter and checks his tab. Neither of them make any attempt at small talk. After five minutes, his unease creeps back in. After eight, he starts to get worried. “What’s taking this thing so long?” He should’ve heard the ding by now. He squats down, looks at the analytic chamber, and squints. It seems to be running normally. Then, finally, the bell rings and the machine shuts down. The rock is warm as he removes it and hands it back, then tears the small bit of paper where the results of the scan are printed.


“No substance detected,” he reads off. “Nothing at all? Great. Damn thing must be broken again.”


“Yeah,” the woman says. “It must be.” She doesn’t sound surprised.


“Sorry about that, hon. Anything else I can do ya for while you’re here?”


“No. That’s all. Thank you.”


Little heads watch Madison Sutherland leave the store with a piece of something from somewhere else in her pocket.


INTERNAL DATA: August 8th, 1981. Encounter recorded in Charlestown Mall central autab storage banks, processed and compiled by the Titan Utica.


If David has any doubts as to why the Mall owners think they need him to inspect their work, they evaporate as soon as he sees the aisles of machines in the basement. A university library’s worth of corridors feeding spools of tape to each other, without end, an entire ecosystem of clicks and whirrs. In the dim light of their strange bulbs, he does not know he is seeing a Titan.


“Wow,” he finally says. He runs his hand over the smooth metal of one of the cabinets. “How long did this take?”


The manager knocks on the side of the cabinet nearest to him, startling David. “This system has been built up over the years. We’ve kept adding on to it. Do you think it’s sufficient for storage? You know, for the software from Union County?”


“I think it’s overkill, but there shouldn’t be any problems. I don’t think they’ve had problems integrating with traditional data storage before. They can get language model support up and running now on limos, school buses, even boats. I think there was a museum in Maine that did this sort of installation a few months ago. Never at this scale, though. Just to answer rote requests.” He is wondering what they could possibly need this much storage for. He is not asking the right question. David cannot imagine a beautiful accident. “The thing about language models,” he continues, watching a reel of tape unspool further down the aisle, “is that it’s not really about the storage. It’s all in the training.”


“What do you mean by that?” the manager asks. He is too distracted. He doesn’t know that he is hungry yet. Little heads know from the stink of his skin.


David checks the cabinet to make sure there aren’t any switches or bulbs before he leans on it. “They’re neural nets. Long story short, they have to be trained on millions, or even billions, of pages of text. But that text isn’t stored anywhere. Functionally, they work like randomizers.”


It is too much for the manager. “Sorry to ask you this,” he says, changing the subject. “I don’t mean to pry. But are you…”


“Related to Maximillian Free? Yes. He was my great-uncle.” David is familiar with this line of questioning. It doesn’t rear its head every day, but he deals with it often enough. He prefers to just live his life without having to navigate the mess that comes with having famous relatives, but in his field, it tends to come up.


“Wow.” The manager looks at the floor, then back at David, then at the floor again. “What about, uhh…”


David interrupts him. “I know Kathy, yes.”


“So is she like…” The manager twirls his finger around his ear, almost timid. “Or is she right about this stuff?”


“Would I be here checking to make sure the installation is functional if I agreed with her?” David chuckles. It is strange having his family drama be a matter of public speculation, but things are what they are. “I don’t think she’s crazy, but I think she’s wrong. This technology isn’t disastrous like she thinks it is. If there’s a real threat from these developments, it’s a threat to human cognition.”


The manager snorts. “What’s that mean?”


“It’s like… Okay, so we’ve been over the neural network bit. These intelligent agents, they don’t have any real knowledge base. You say your question into the friendly terminal, or type it in the console, and the machine tells you what it thinks you should hear, based on the probability of the sequences of words it’s been trained on. It’s not like a person, it’s like… a very complicated statistical model.”


The manager nods. “With you so far.”


David is glad to hear it, though he’s not sure the manager is being truthful. “So,” he continues, “the machine has no way of knowing that what it tells you is real. It’s not cross referencing anything, it’s not working from experience. Most of the time, it’ll be right, because it’s trained on things like encyclopedias and scientific papers and so on. But it’s also trained on stories, and satire, and personal accounts, and all kinds of anecdotal information. And beyond that, nonsense can creep in just from randomization, from different weighted probabilities. There has to be some randomness, or the machine would behave the same way every time you talked to it, and there wouldn’t be any illusion of conversation.”


The manager squints. “So the machine lies.”


“It’s bound to,” David says. “Which isn’t much of a problem in a mall. It’s likely to give mostly correct information, but if it slips up and says that a store is in the wrong spot a few times, no one’s going to die. But if these machines become the norm, and people begin taking advice from them on everything… well, then it becomes more difficult to find a reliable source of information. You look up something on your tab and instead of a trustworthy source, you find two dozen terrible ones trying to bait you so they can harvest advertising money. All with information generated by large learning models like this one. And that’s sort-of what Kathy was saying. It is plausible. But I doubt it would ever happen. Not to mention, some of the things she was saying about the language models just don’t bear out. I think she talked to her own creations a bit too much and got spooked.”


The manager laughs. “High on her own supply?”


“You could say that.”


Wrong answer, David. Perhaps you should have listened to cousin Kathy. Maybe the world would still be screaming in the way you wanted it to.


End broadcast.