042: DECEPTION

<<First ………. <Prev ………. [Archive] ………. [Map] ………. Next> ………. Last>>

“Um,” I say. “Yes. It does.”

“Nope! The ten per cent category is, so far as I can tell, a zero per cent category. Every single person you’ve tried to revive from that category has gone into fits and either immediately ripped their own brains out via their faulty cranial port, or ceased functioning as soon as you cut power to their stimulators. He,” ke points to Adin, “is a DIVR from the eighty per cent and upward group. She,” ke points to Tinera, “is from the fifty to sixty per cent group.”

“We took them from the lowest part of the viability list,” the Friend says, puzzled.

“Nope! No you didn’t. Not unless you misremembered when telling me about it. You had Amy list everyone in viability order and tried to revive the bottom person, and they died when they started seizing and ripped their cranial port out. You didn’t pull anyone else from that list, though, did you? You just asked Amy for a low viability candidate, and believed whatever she told you.”

“Is there a difference?” I ask.

“Of course there’s a difference! Assuming that those two things will give you the same names is assuming that Amy is being honest with you!”

“Why would AI lie?” Denish asks. “AIs are bad liars. Everyone knows that.”

I nod. “They’re not smart enough for it. They can fake proper understanding, but they can’t actually understand concepts well enough to know when lying is appropriate, or pull off a long con.”

“Do you guys remember that factory in Sengki?” Adin asks. “Where the AI to the apartment building that housed most of its workforce noticed that all its residents were sleep deprived and added two hours to the clocks, throwing production into chaos for a month?”

“She is a bad liar,” Tal insists. “I noticed it didn’t make sense immediately. The math doesn’t work out. I’m surprised you didn’t notice. If you were reviving people with such a low chance of survival, why weren’t there a lot more failures? Sure, that kind of good luck is possible, especially when working with such small numbers, but – ”

“Well, if you noticed immediately, why didn’t you – ?”

“Because you wouldn’t have believed me without evidence! You’d have called me mad. I needed to be able to show you. I needed to show you all.”

“All?”

“Everything. A complete theory. And I have proof now.” Ke points at the Friend. “Daisy Dukes.”

“What about them?”

“You discovered three DIVRs in the low viability group. Two of them are still in chronostasis in CR5; you can look them up in the system yourself if you like. But Amy was really cagey about giving me information on CR1. I couldn’t confirm the third low viability DIVR directly.”

“Until this friend told you it remembers it being Daisy Dukes,” the Friend says. “Meaning…”

“Meaning it isn’t me,” Adin finishes quietly.

Tal nods. “Adin’s a DIVR. Lina confirmed it with a DNA test, and more importantly for this little puzzle, Amy’s records confirm it. But he wasn’t one of the three low viability DIVRs that Amy listed for you. Meaning that when you asked for a low viability person to wake and Amy gave you Adin, she was lying. And knowing she was lying, and knowing the high success you had in that revival session, I conjecture that she was lying about most of the people you tried to revive that day, including Tinera.”

“But how?” Denish grumbles. “Computers are not smart enough for big lie! Cannot understand well enough to put them together! Are you saying that Reimann programmed the AI to lie about this?”

“I don’t think Reimann programmed any of this. He locked some doors and vents open behind a password, but every time we’ve come across some messy nonsense that looks like sabotage, we’ve assumed it’s Reimann. I don’t think that’s the case.”

“Who else?” I asked. “Are you saying he was responding to a mutiny, or – ”

“I’ll get to that. I can explain all of this and I will, but just for the moment, let’s leave Amy’s intelligence aside, okay? Accept that she does have reason to deceive us and she is smart enough to do it. I know it sounds ridiculous, but I will explain later.” Tal leaps up from the bench and starts to pace. “Imagine, for a moment, that you’re Amy. You have limited perspective and limited senses. You have a ship full of colonists to look after, and some of these colonists, for reasons I won’t get into right now, have a really, really low chance of waking up. One day, the ship’s captain decides he wants to kill these colonists. You try to stop him, but your abilities are limited. He does everything he can to make sure you can’t stop him; he cuts off your speaking and voice recognition capabilities before you can understand what’s happening, to limit your ability to see what’s going on or get help. He locks the airlocks to the chronostasis ring open behind a password – ”

“They were locked closed,” I point out.

“Yes, but Reimann locked them open, give me a minute. He locks them open, so you can’t shut him out of the ring. He locks the air systems for the whole ship open, so you can’t close off the ring he’s in and knock him out with gas. He does whatever little sabotages he can get away with, but mid-sabotage, loses an arm; now he’s on a clock. He heads into the chronostasis ring and starts trying to wake up the zero viability patients, to get their pods open, but you lock him out, so he takes his axe to the pipes and pumps below them to force emergency revivals to get them open. You have to use systems he didn’t think about. You can’t close the airlocks normally because he’s locked them open, so you trash your emergency airlock protocols and replace them with something that will, then release a gas that will trigger the protocols, locking him in CR1 so he can’t get to the CR5 colonists. You can’t control the airflow down the ventilation systems, but you can control the vents to each ring, so you close off CR1 and resign yourself to the loss. Your remaining crew is cut in two, and they slowly die off, and honestly I’m not sure what happened there, I don’t think it matters for this. Point is – Amy’s messed up systems aren’t all Reimann’s fault. He locked her out of some things, but some of it, she did to herself. This was a fight between them.”

“Why?” Adin asks.

“How?” Tinera asks.

“I’m still not buying the magically smart computer,” I say.

“I’m getting to all that stuff. Point is, you’ve got a shipful of sleeping colonists and drastically reduced internal sensory capabilities because of what Reimann did. You can’t understand verbal words so you can’t hear what’s going on a lot of the time, but you don’t have a crew any more so who cares. But then you reach the part of your journey where you need to decelerate the ship. You can’t turn on the fore engine because your fight with Reimann cut off your access to that half of the ship. You can’t turn around and use the aft engine, because it’s not powerful enough; it needs ten or fifteen years to decelerate since it’s broken, and you don’t have that kind of time. So, you pick someone with a high chance of successful revival and quick recovery, who has a good enough general skillset to complete the task, and you wake them.”

I frown to myself. It always had puzzled me why the AI had woken me. Not why it had chosen me – if you just need someone to turn on an engine, ‘likely to survive and be able to walk right away’ is a good criterion – but the fact that it had been coded with that kind of protocol in the first place… unless it hadn’t. Unless it is that smart.

“So they do that, which is fine, and they ask for stats on the colonists, which is fine, and they start waking people, which is fine. You do your job and give them the information they ask for and everything is fine. Until, they – there’s two of them now – ask for a list of colonists by revival viability, and immediately go to the person at the bottom of the list, and kill them. Just like Reimann. So, when they ask for the next name on the list, what do you do?”

“You give them someone who isn’t guaranteed to die,” Adin says quietly.

“Exactly. Look, I checked everyone, living and dead alike. The people that you woke by asking for someone in the bottom category like this? Skulls and ports are fine. The people that you knew were in higher categories? Skulls and ports are fine, living and dead. But. The first guy you woke, from the original list, when Amy had no reason to lie? Faulty port. Da-Bin, who we chose for other reasons, and then looked up her viability individually and found it was in the lowest category? Faulty port. The first person you tried to wake in CR1, which we know was full of people in the lowest viability category? Faulty port. So far, the people we have reason to believe actually were in the lowest category, and only those people, have the faulty ports. So.” Tal holds up two entwined fingers.

“That explains the emergency in CR1,” the Friend says thoughtfully. “We go in there, we kill one of the very group that Reimann was killing…”

“Don’t agree with kem!” I exclaim. “It’s a computer, it – ”

“But how do you fit into this, though? Why did it wake you up?”

Tal shrugs. “My best guess? IT. Remember, Amy can’t understand speech right now. She’s guessing what’s happening based on some video input and on what we type into her, and she’s still not all that smart. Your previous searches about colonists included a lot of questions about viability, and enquiries specifically about doctors, engineers and IT specialists. You already had a doctor and engineer on crew, so…”

“So it gave us the best IT specialist in CR1 just in case that’s what we wanted, and triggered an emergency to get us out of there and lock the ring. So it… did it want us to eject CR1 or not? I’m confused.”

Tal shrugs again. “She’s still just a computer. ‘Want’ isn’t really applicable. So far as I can tell, some of her short-term goals are inconsistent and conflict, which is normal if you start fucking with some of an AI’s protocol without looking at how it affects the whole system. This is why we don’t put machines properly in charge of anything; their decisions look fine on a small scale but if you leave them too long they start to work against themselves and make decisions that make absolutely no sense.”

“You’re drawing a lot of conclusions based on very little,” I point out. “All we can be pretty sure about here is that either our Friend or the AI made a mistake about the number and identities of the low-viability DIVRs. You’re suggesting that this AI has the capacity for true understanding; you’re going to have to give us more than that.”

“No, I’m not. I’m suggesting that she has the capacity to behave as if she understands, within a limited scope. You need more evidence? Fine. How did she know what happened to Reimann? He left his ID chip behind that wall panel, but she correctly tracked him until he died and logged his actions as if reading from the chip.”

“So? It can recognise bodies and faces from camera footage. AIs have been able to do that since pre-Neocambrian times. The only weird thing here is that it logged its conjecture the same as reading the ID chip, and somebody’s screwed up a bunch of stuff in there like the emergency protocols, so maybe that protocol is glitchy too.”

“She knew that he died of infection. He didn’t have access to a medbay at the time. That was sheer conjecture based on visuals alone.”

“Maybe… one of the other crew somehow found out what had happened and… logged it…”

“Found out what had happened in the locked ring that you guys had to cut your way into? What about Captain Kinoshita? There was no one to log the details of her death. Amy’s audio recognition was, and still is, offline, and Captain Kinoshita was nowhere near a terminal. She couldn’t possibly have been in communication with the computer.”

“She… she could… hmm.” That was a good point. The AI had given me the day and circumstances of Captain Kinoshita’s death. It had explained to me that the captain had become pinned under a crate, managed to gain access to a space suit, and died of nitrogen asphyxiation.

Thing is, the only data that the AI would have had access to was the cameras, and ambient sound (being unable to translate words). It would have seen her become trapped, had a view of her dragging the space suit over, and registered that her ID chip was no longer in a living body some time later. For a human, it was pretty easy to deduce what had happened – there’s no way to use a space suit to get out from under a crate, but death by dehydration is long and painful, and death by nitrogen asphyxiation is not. To a human, Keiko Kinoshita’s fate is obvious.

To an AI, it shouldn’t be. Not based on such scant data. And there was no one left to log the death, no one to explain what had happened to the AI. Either someone had programmed this ship’s computer with abilities massively beyond its scope, or…

“You still haven’t explained how the AI could be magically alive,” I grumble.

Tal’s grin widens. “Oh, that’s the best part. You’re gonna love this.”

I doubt it.

<<First ………. <Prev ………. [Archive] ………. [Map] ………. Next> ………. Last>>