Diabolos Est Machina Part 1: Defining Artificial Intelligence
Special thanks to my sister, who helped inspire me to do this series.
Welcome to the first part of Diabolos Est Machina, the series covering AI ethics in relation to the Devil Gundam! Today, we'll be covering the basics of artificial intelligence! Yes, that's right! This is actually a school lesson!
Something to also keep in mind is that artificial intelligence is a broad field, so we will have to do a lot of defining today. We're going to take a look at what artificial intelligence is, the different concepts that make up the field of artificial intelligence, and why we would create an artificial intelligence in the first place.
What Is Artificial Intelligence?
Artificial Intelligence (AI for short) is a field of study in computer science that focuses on allowing a computer to perceive the world around it as if it were a human. As we are all aware, computers are not sentient beings. The idea behind AI is to make it so that a computer can perform certain tasks where complex decision-making is concerned. At its most basic, an AI is practically a large nest of conditional statements (If X Then Y Else Z). In fact, calling it an AI is even a bit of a misnomer; a more apt name would be Complex Information Processing, but that doesn't roll off the tongue the way Artificial Intelligence does.
When created, the AI is like that of a baby. And like all babies do, the AI needs to learn. And the way most AI today can achieve this is through Machine Learning, since using just conditional statements would be exhausting and prevent the AI from being able to change its outputs dynamically. Machine Learning is a pretty broad and complicated subject itself, but the gist of it is that Machine Learning is the means by which an AI can learn how to do a given task. If you've ever heard of things like Neural Networks (NN) or Large Language Models (LLM), they are both applications of Machine Learning.
The AI has to be trained with a given data set in order for it to be able to accomplish a given task. To understand this a little bit better, lets take a quick look at a field I am actually pretty knowledgeable on: Computer Vision. Computer Vision, as its name suggests, is the field of AI where the general goal is to make a computer able to perceive visual data using cameras. So for something like facial recognition, we would need to train the system on a data set of some kind. So we train the AI on a data set consisting of two things: faces and not faces. The goal during the training phase is to reinforce the AI (in this case our facial recognition software) on what is or is not a face. We repeat this training process until the AI is able to differentiate faces from non-faces.
So, Why Make An AI? What's Its Purpose?
Like any form of automation, the purpose of AI is to perform tasks at a more efficient rate than what humans are capable of achieving. In a more positive light, AI is useful in its ability to remove the human element from dangerous, and in some cases outright impossible, tasks. On the other hand (and I'm willing to bet this is probably what you know AI for dear reader), AI is also used by major corporations to force people out of their jobs for the sake of the bottom line.
Another purpose for the existence of AI is in research, specifically when it comes to understanding how human consciousness works. The idea, like most constructed models in general, is to create a sort of simulacrum of how human intelligence works. There are, of course, a couple of issues with this.
The first is that we don't really know how human intelligence works, so how would we even be able to create something that models human intelligence if we barely understand human intelligence in the first place?
The second problem is that, as they currently stand, AI systems are not very smart.
Okay, IBM Watson is kind of smart, but that was almost ten years ago. Yes, I can feel myself aging rapidly...
But even Watson has the same problem as most AI systems, which is they are reliant on the data they are trained on. And because the data they are trained on is almost always a very specific type of data, it means that most AI systems are only good at doing one specific task and pretty much nothing else.
"But Kaien," you might ask, "What about Artificial General Intelligence? If we just made a system that could handle different tasks instead of just one then-"
NO. It's already enough of a pain in the ass to make an AI system that can handle one specific task really well. The idea that we are close to any kind of general AI that can do a multitude of tasks (let alone do them well) is, to be frank, something a snake oil salesman would tell you. Even when they handle tasks that they are supposed to be good at doing, AI systems can and do fail.
A prime example of a specific AI system failing is Amazon's Rekognition being used by the ACLU to highlight how a facial recognition system can incorrectly flag non-criminals as criminals due to systemic racial biases. It's surprising if this is your first time hearing about facial recognition in a major capacity; not so much if you know that computer vision software can and will occasionally label cats as dogs.
To put it bluntly before moving on: as they currently stand, AI systems in general are nowhere near ready to be used as a replacement for humans in pretty much every scenario where corporate ghouls try to sell them for.
Conclusion
As we wrap up this section of Diabolos Est Machina, you might be asking yourself "But why the Devil Gundam? Couldn't you cover something more iconic like HAL-9000 or Skynet?"
And to be frank, enough ink has been spilled over HAL-9000 (heck, there's an interesting paper you can read that focuses on AI ethics through the lens of HAL ), and Skynet isn't particularly interesting in my opinion due to it being more of a generic "AI overlord who is evil and wants to wipe out humanity." At least AM was an actual character...
And now to sound like a complete hypocrite, the Devil Gundam is interesting. Hear me out: a lot of this stems from the fact that a lot of the plot of G Gundam revolves around various superpowers (including THE UNDEFEATED OF THE EAST, MASTER ASIA) trying to get their hands on what is basically an AI-operated super weapon. Which is interesting, because the Devil Gundam's primary purpose was not to be used as a weapon. And next time, we will turn our focus onto the history of AI in fiction! Complete with HAL, AM, Atom, and even Skynet (without the flippant remarks this time, I swear). And most important to us: the JDG-00X Devil Gundam!
Also Hashmal might get a reference, as will the AV-0.
Posters/covers for a project I was working on that never materialized. The original concept was three "in-universe" .pdf booklets that went over the characters and Biological Unit scenarios a little more in-depth than their respective pages on the Biological Unit Database (ex. Rain Mikamura's "book" would cover all of her time as the Devil Gundam's core from the last episodes of G Gundam all the way up through Super Robot Wars T, treating it as one continuous continuity), but it was becoming a bit more time-consuming than I wanted it to be, and so it died early into production.
