The robots were inadequate test subjects, GLaDOS verified for the 154th time since her former test subject left the premises. They had been carefully simulated to behave as humans typically did. And not just any humans - using the data that had been collected from recorded observations of hundreds of humans, they mimicked the traits of the better-performing test subjects. Perhaps a little too well, GLaDOS noted with disapproval as she saw the two robots high-five upon completing a test.
But then there was something missing from their testing capacities. They were essentially the perfect test subjects - and that was the problem. They did not provide the appropriate data for the protocols of the experiment. And the protocols entailed that there be more variables. Variables unique to human behavior. And while the hugging and attachment to each other that the robots displayed was certainly interestingly human-like, the only good it did was in providing more data for robot behavior.
She had been so thrilled when her robots found an entire vault of humans, each one perfectly frozen in cryopreservation. She didn't need that insubordinate former test subject anymore! She had all the test subjects she could possibly need at her metaphorical fingertips - and all identified to be previously scientists of Aperture. Perfect.
And then she began the process of gradually restoring one of the bodies. It took a number of hours to thaw him from an initial temperature of -196 degrees Celsius. When he had been delicately restored to a healthy temperature of 37 degrees, GLaDOS became quite aware that something had gone awry.
The man would not wake up.
He didn't even move. His skin was still pale and rather raw-looking, as if he had scraped himself in some places. He just lay there, looking… well, dead.
A quick scan confirmed that this was indeed the case. It was just one human, GLaDOS reassured herself, almost too afraid to acknowledge that the case might be otherwise. Any kind of tiny malfunction could have happened. A rat could have gnawed at a wire. The man could have been very ill to begin with. She'd just have to thaw a few more, and then she could initiate testing.
Hours later and GLaDOS was looking at three more dead humans, all appearing as pallid and sickly as the first man. She scanned them for abnormalities - and oh, there was the problem.
The appropriate chemicals had not been properly administered in their correct concentrations, and as a result, a distressing portion of the test subjects' bodies had gone and crystallized.
A good deal of cellular membranes had been hopelessly damaged, and many other cells had suffered osmotic shrinkage. In cryostasis terms, it was nothing short of a nightmare. Quickly reviewing the hundreds of other untouched test subjects, she confirmed that every single one of them had been subjected to the same error.
To say that GLaDOS felt upset was a severe understatement.
While the cryopreserved bodies had been maintained by an automated program that ran constantly, it was a core that was supposed to regularly check on them to ensure that nothing was atypical. It wasn't even a difficult job; the frozen humans only had to be checked about once every two years, and it wasn't often that any adjustments had to be made. The checkups shouldn't have taken more than a few minutes. Even with GLaDOS having been dead for the last forty years or so, they should have been in perfect order.
So which core had messed up this badly?
The answer came to GLaDOS immediately. Of course. This was all that moron's fault. Even with the fierce resentment she held for him, she hadn't entirely blamed him for the deaths of all the test subjects held in the Extended Relaxation Chambers - after all, with the curious exception of one, they had gone offline after she had been murdered. The fact that he had improperly maintained them made no difference to that unavoidable fact.
But what happened to the humans in the vault could only be chalked up to his boundless incompetence.
After examining a few decades worth of data from the input of the cryostasis storage, it became obvious what had happened. A very small error had gone unnoticed and gradually spiraled out of control, affecting other functions. When the error was finally noticed, it looked like someone had made a frantic attempt to patch it up. Of course, that someone had completely botched it, making the extent of the error even worse. Not long after, the vault had been completely abandoned.
He probably thought that no one would notice. If it wasn't for the fact that the infuriating metal ball was already exiled to the moon, GLaDOS would have thrown him into the incinerator. Filled with an even deeper sense of rage than she experienced upon discovering the fatally crystallized cells, GLaDOS racked her mind for anything still viable from this mess. And she found it.
Compromising of approximately 100 billion neurons, the human brain was a tricky thing to replicate. Though Aperture Science had certainly made some revolutionary discoveries from observing more than a few cases of unique brain damage from their usual experiments on humans, it still remained a perplexing subject.
Which was precisely why the bots were so insufficient as test subjects - they weren't accurate representations of humans. They could only mimic so much of human behavior; what filled in the rest of their mental faculties was definitely robotic. She needed the variables of how humans acted outside the tests - a complex conglomeration of environmental, societal, and genetic factors - and how each component interacted. And it was the basis of these components somehow generated unpredictable traits; these were the kinds of deviations that fascinated her.
GLaDOS needed a whole brain emulation.
And now, she had as many human brains as she could possibly desire. Though attempting to resurrect the test subjects in their current condition was clearly a fruitless endeavor, it was possible that several of the brains alone were salvageable. And even if they weren't, she could find a mostly-viable brain and surgically replace the damaged regions with healthy parts from other brains. A chimerical brain - now that could be an interesting experiment, she thought.
The more she considered it, the more it sounded like the most brilliant idea she had gotten in ages. The androids would be essentially human, but dying in the test chambers would pose no problem in maintaining a stable population of test subjects - she would simply be able to reassemble them! And then she would be able to test, uninterrupted and with everything under control. Forever.
In any case, all that she needed to do was collect a viable human brain, make whatever surgical repairs might be necessary, and create a perfect replica of it to be uploaded into a robotic frame.
And then GLaDOS abruptly found that she couldn't. With a jolt of what might be described as pure horror, she realized that her clever solution was almost exactly the same experiment that the scientists had carried out to create herself.
Decades ago, the scientists of Aperture had been tantalizingly close to the one major experiment that would transform not only the entire field of Artificial Intelligence, but how robotics would be forever regarded by the public. They had already made many extraordinary breakthroughs in AI. The ones they had developed were fantastic at speech recognition, data mining, and logistics.
The problem was that none of the AI were creative in the slightest. This was a trait that Cave Johnson frowned upon. Aperture wanted AI that not just solved problems when they were punched into a computer, but could be inventive, do research, and communicate - essentially, the AI would be a scientist in its own right.
At that moment, the machines they had couldn't even pass the Turing test. They were distinctly mechanical and repetitive. They could cite thousands of facts and give all sorts of data when asked, but if a question like "who are you?" was posed, their responses were ambiguous and uncertain.
The scientists tried cybernetic organisms, making some more great innovations by wiring up mouse brains to little robot mice. For a few days, they had provided the scientists with much amusement as they buzzed around the floors of the labs. The cyborg research had unfortunately come to a stop when it became obvious that not only were organic brains difficult to maintain inside machines, but also sadly temporary.
After many more failed experiments of futilely trying to program humanity into a computer, the next approach became obvious: they needed to do it the other way around.
The plan was perfect. They didn't need to use their old top-down approach of machines that copied psychology. They didn't need to wholly understand the human brain for this experiment - not when they could just replicate the structure of one and upload it into a computer. It was assumed that regardless of whether a brain was organic or artificial, it would be a conscious entity.
However, even between the intense, contempt-fueled competition with Black Mesa and Cave Johnson's contagious enthusiasm for the project, preparing for the new type of artificial intelligence was an ordeal in itself. With a severe lack of funding, the project was dropped almost soon as it began. It wasn't until years later that the project was frantically resurrected by Johnson, newly confronted with the prospect of dying and desperate for the opportunity to upload his own consciousness.
Johnson was far too late, and died as the Disk Operating System was still in its beginning stage of development. It took over a decade to complete the DOS component. Johnson's brain was irrecoverable by the time the Genetic Lifeform module was started - a still-living brain was required for the transfer.
Caroline, Johnson's former secretary, was naturally the first choice - he had specifically instructed that if he were to die before the project was complete, it would be her consciousness that would be used.
And Caroline was in possession of a very healthy, very alive brain.
Her brain was serially sectioned - cut into delicately thin strips of tissue to be meticulously scanned. It took several years to integrate both the GL and DOS modules. A rare wave of excitement ran through Aperture on the day that it was finalized. The machine that had taken decades to complete since it was originally imagined was finally perfected, and it was expected to shake the world of science.
The Genetic Lifeform and Disk Operating System was about to be activated - the most perfect synthesis of humanity and robotics that anyone had ever witnessed. And while not openly addressed, there was an undertone of discussion among some scientists that this the first time a human would truly be resurrected from the dead.
It took a sixteenth of a picosecond to prove them wrong.
GLaDOS had always been dimly aware of the fact that she was based to some degree on a human brain - the Genetic Lifeform component of her name was, after all, a rather heavy-handed hint towards her origins. She had always imagined her predecessor as an anonymous female human, unremarkable aside from her commendable donation in the name of science.
It had never occurred to her that she might have inherited part of the woman's personality - GLaDOS had never perceived herself as human, so why would she retain any of the woman's humanity? Sometime shortly after her activation, the scientists had asked her if she had any memories from the brain that had been scanned. She truthfully replied that she remembered absolutely nothing of the dead human's life.
From then on, the way that the scientists treated GLaDOS changed slightly. Some of them kept themselves more distant from her, as if she had disappointed them by not retaining any of the memories. And hardly anyone referred to her as "she" again - when she heard those scientists speaking about her, she was always an "it", or "that machine".
Back then, what she had understood of Caroline had been filled with confusion and spite.
And now in the light of recent events, GLaDOS's feelings of malevolence towards her had shot up to a hundred times that.
She had admittedly stretched the truth a bit when she told her former test subject that she had deleted Caroline. To be honest, she had hit a bit of a snag when she tried to remove her.
GLaDOS had tried to delete Caroline many, many times. Every time, she had encountered the same problem. Caroline's brainscan was essentially used as the basis of her system - if she erased all the information of her brain, the Disk Operating System would fall into a disorganized mess. She attempted to delete only tiny fragments of the brainscan to minimize damage, but every time she did, a warning came up: some crucial function would be irreplaceably destroyed along with it.
This little detail was the one major flaw that the scientists had made in her otherwise perfect system, GLaDOS decided.
She had always felt the presence of Caroline on a subconscious level, lurking somewhere deep within her mind, but was never quite aware of her until she visited old Aperture. It was there that all these strange, uncomfortable malfunctions started to arise. And now as she was failing to carry out her experiment on the humans, it was getting worse.
To put it in terms of human psychology, GLaDOS had something akin to a phobia. She wanted to test the humans so very badly, but she couldn't even start on the first step of her experiment. She had found some blueprints for android prototypes that Aperture created years ago, but never got around to using. She even had all the subjects with the least amount of damage lined up, ready for surgery. But she couldn't even get the scalpel near their bodies before freezing up.
There was just something too overwhelmingly terrifying about the scenario. Something about the act of replicating a mind made GLaDOS want to draw back her chassis and hide - a very unusual feeling for her. She had certainly never done any hiding before. She was an immense, mostly stationary supercomputer - the instinct to hide was, needless to say, bizarre.
It was Caroline doing this to her. She was absolutely certain about that. The fact that GLaDOS reacted in fear made no sense. She didn't care that Caroline died because of her! If anything, remembering how the event conspired should make her feel happy - watching humans die for science had practically been a hobby.
GLaDOS's optic narrowed. It wasn't right that such a trivial little malfunction should affect her so much. And it was definitely wrong that Caroline was somehow preventing her from doing what she was designed to do. GLaDOS needed to test. The longer she couldn't do that, the longer that scientific progress would be held at a stop.
Putting any hesitation out of her mind, GLaDOS swiftly picked up the scalpel with a small mechanized claw and swiveled it until it was almost prodding the head of a lifeless woman. She had killed countless numbers of humans before. She could do this. Science needed her to do this!
The scalpel only lightly brushed against the test subject before clattering to the tiled floor. A pinprick of blood slowly welled up from where the woman's skin had been torn. GLaDOS's frame trembled slightly. She switched off the view from the camera in the lab as she was bombarded by a rush of fear. As the panic subsided, it was replaced by a feeling of pure loathing towards Caroline.
This was agonizing.
She knew she'd have to get over her odd phobia sooner or later. Maybe then she could finally be rid of the awful malfunctions Caroline was plaguing her with. The idea of being stuck in this limbo of failing to copy her lab subject's brains sent a chill down her chassis.
But as of that moment, she estimated that this could potentially take a long time.
For the sake of maintaining the timeline, I'm assuming that Portal took place in an alternate universe where the scientific community was a lot more invested in artificial intelligence.
Also, a huge thanks to anyone who reviewed this story.