The scenario described below is for a role-playing game. The setting is Earth-like but with a healthy dose of sci-fi. Even though the setting is sci-fi I would prefer answers be largely based in real science (no un-obtainium based answers).
Background:
In my world humans coexist with hyper intelligent AI. Although the planet is Earth (or at least very Earth-like) the actual origin of these AI is unknown and they were not created by humans. These AI beings are composed of two separate but linked components:
- Machine Intelligence: a vast analytical mind that functions in ways that are impossible for human beings to understand. The AI can calculate and comprehend many things that are out side of human understanding. However, the Machine Intelligence has very little in the way of goals, drive, or ambition. It is largely content to observe and process data.
- Human-designed Personality: humans have designed a way to control and interface with the machine intelligence. This programmed layer sits on top of the Machine Intelligence and provides a human-like personality for humans to interact with and organizes the Machine Intelligence's processing power to tackle computing problems of interest to humanity.
AI using a combination of their innate Machine Intelligence coupled with a Human-designed Personality are employed to monitor and control many aspects of human existence (from advanced forms of modern smart home devices to AI air traffic controllers).
From time-to-time there have been instances of a AI's Machine Intelligence and Human-designed Personality becoming uncoupled/damaged. The results of this uncoupling can be catastrophic if the AI was being employed to perform a major task but it is never malicious. The aforementioned AI air traffic controller could kill many humans if it suddenly lost interest in its job, but it would not be actively trying to harm humans.
The Scenario:
I have a scenario where an uncoupled AI appears to be murdering humans. The AI (using robots, devices it controls, etc) is killing humans and harvesting some material from them. This horror scenario appears to be the world to be the first instance of an AI with ill-will towards humanity.
In reality we have some variation of the Paperclip Maximizer (Instrumental Convergence). Some incompetent (or malicious) programmer coded a task/imperative into the Machine Intelligence itself and now the AI is continuing to attempt to complete that task without any regard to humanity and their safety.
My Question
What task is my AI attempting to carry out and why does it appear to be selectively targeting humans? Ideally the AI needs to harvest an actual physical substance from humanity and that physical substance is found semi-exclusively (or is simply readily available) in humanity.
Traits of desired answers:
- A specific substance that the AI is attempting to harvest from humanity.
- The chosen substance is difficult to find in other living creatures (creating the impression that it is specifically violent towards humans).
- The substance is impossible/difficult to find in the environment (at least in the same form it is found in humanity).
- Bonus points: The AI is exhibiting some other strange behavior that could provide a clue as to the real reason it is killing humans. This secondary behavior should either be a second source for the substance or possibly a second substance used in the processing of the substance extracted from humanity. Ex: the AI is also harvesting only the peels from bananas.