A case against AI in game development from Chet Faliszek (Half-Life, Portal, Left 4 Dead, etc.)

TL;DR: Chet Faliszek, writer of Half-Life, Portal, and Left 4 Dead, uses the space station game he is developing in Godot to explain why you should not let AI make games for you—the process of getting hands-on and researching real-world reference material is itself the source of creativity, writing, and design. ## A Starting Point: A Space Station in Earth Orbit Chet Faliszek is making a game in Godot: the player is alone on a space station above Earth, communicating with the ground. He will share more details later, but that is his starting point. His initial idea was small—to make a very small project in Godot. A space station is actually quite small and can be loaded entirely into memory, so he could keep things simple. It is a heavily story-focused game because Chet wants to write a lot. There are characters on the ground, there are characters, and then there is you—the player. ## What Is the Gameplay? Making “The Work Itself Is Annoying” Part of the Experience Even with a story, players need to stay active. They cannot just click through dialogue trees. He was looking at and playing some newer space games at the time, and their approaches generally fell into two categories: in one, as soon as you interact with a screen, everything happens on that screen, putting you on a 2D plane where you spend most of your playtime; the other is 3D, letting you float around inside. He did not want to deal with either, nor did he want to make movement complicated, because that is a challenge in itself. He wanted the game to be about simple movement and the nature of everyday life—you know what to do; it is just work. And actually going and doing that work is really annoying. That was the feeling he wanted to capture. He particularly dislikes the interface controls in some games: once you enter 2D, you feel trapped in the screen, with “a mouse inside a mouse inside another mouse.” Operating and moving around those interfaces frustrates him enormously. ## Back to the Source: Apollo Missions and NASA Panels “Hey, I’m old. I remember these guys. Look at these guys. We will never be this cool.” These photos came from the Apollo missions, from NASA. Back then, they were simply cooler than us. Honestly, that is just how it was. He started looking through old concept art. For example, he thought the concept art for Space: 1999 would make him go, “This will blow me away,” but instead it was just bland and boring. He wanted something cooler. He noticed the details of the displays on the Apollo missions: in the photos, they did not look like cathode-ray tubes—no flicker, no scan lines running from top to bottom. That was because they used something else. ## Charactertron: A Character Display It was called a Charactertron, a character display, and it actually worked by passing light through it. That was how those things were laid out and configured. There was a time when people would say, “NASA is way ahead of us in everyday technology,” and this was what they meant, because NASA had all those incredibly smart people working there. Chet really likes this, but he does not intend to recreate it. It would still be on a screen. ## One Button Does One Thing NASA also had many panels that followed one principle: one button does one thing. He showed a picture of a communications control panel with dedicated volume controls. He particularly liked the rotary-dial telephone mounted at the top. Speaking of rotary phones, he made a joke: back then, when you came home drunk from a bar in college and she had told you, “Hey, call me when you get home,” if her number had zeros in it—that was a test. The point is that all these panels had separate, individual functions. They did not make one button do too many things. He really liked that. His game is not a recreation of this and does not need to be historically accurate, but it can be influenced by it. This is the concept design process. Usually, he would have a concept artist, and the two of them would discuss things and bounce ideas off each other. Here, though, he had to go back to the source. ## Why Not Let ChatGPT Build a Spaceship? “Why not just have ChatGPT build me a spaceship? Make it look modern and cool?” No. Because he is tracing things back, learning, and remembering them. Every one of these things inspires him. What is cool about NASA reference material is that there is an enormous amount of imagery. He thought of Apollo 13—Tom Hanks is so good in that movie—the part where they stir the tanks, an explosion happens, and then they have to get home. That switch is really cool, so it will end up playing a role in the game. ## Learning to Make Knobs in Blender Meanwhile, he is learning Blender: knobs, knobs, knobs. He wants a knob; he wants physical controls—if you are not operating things on a screen, you need physical controls. Although he has to reuse buttons, he needs some things that can cycle through settings, because he wants to make adjustments, such as going from zero to one between two things. He spotted the knobs on an old oscilloscope from that era. Looking down at them, you can see that they actually saved design time by making one knob—the most complicated one—that essentially contained all the other knobs within it. If you made that knob, you could effectively break it down into a whole bunch of knobs. Chet says he is really a beginner at Blender. His trick is learning to break things into small pieces so that adjusting one part does not mess up everything else, along with things like beveling. He does not care what people say is the right or wrong way to do things in Blender, because now he knows how to use it, he can make things, and he likes that. ## A Real Switch, and the Line of Dialogue It Inspired The switch he wants to make is something you can actually buy in real life. The reference photo came from a real object. Originally, he wanted something with an ’80s science-fiction feel: a channel selector with a screen above it that you keep clicking because it is searching for the next channel. But if you have never used a channel selector, there is a problem—for example, if that channel has gone off the air late at night, your clicking range gets smaller. He would use an old cathode-ray tube for that kind of screen with scan lines. He wondered whether some of those buttons should be adjustable now, “but that might be too much.” But he would rather return to that real switch, because it will eventually become a line of dialogue in the game. What he is doing is mixing and matching: those single-function buttons from ground control panels, combined with things from the space station. This button comes from the space station. It has a scratch on it because they were wearing gloves at the time and wanted to make sure they would not accidentally touch something else while wearing thick gloves. This is not the kind of switch in a fighter-jet game where you flip it up and the bomb underneath drops. From a design perspective, that raises a problem: when you click it again, are you flipping it up or activating it? It is much more sensitive than you might think. So we set those aside and use this. ## The Environment Tells the Story: Why Am I Not Wearing Gloves? But my character is not wearing gloves. And my character is a novice too, so he does not understand what this is and asks questions about it: “Wait, why am I not wearing gloves?” This helped him realize that it is not Chet speaking—it is the environment telling the story, the character interacting with the environment, which is exactly what he loves. The answer is that you are on a space station, and they do not actually wear gloves inside a space station; but it is also because you are just going through the motions. You have only just arrived, you have no business being there, and so you do not have space gear. Meanwhile, the people on the ground are trying to explain this to him more gently. The character can then come to understand: you are not wearing gloves because we do not particularly care what happens to you—you do not need a spacesuit—and B, you are not actually doing anything, man. You are just up there so that someone is up there. Beyond that, we do not care about you. Simply having this button allows Chet to discuss these things in a less explicit way. Through this button, players learn that they do not understand the system they are in; that the system was cobbled together from parts used both on the ground and in space, hastily assembled for various reasons; and that—“Hey, just stay up there. Shut up.” So his conclusion is to find a button that actually existed and use it as source material for concept design, instead of going to ChatGPT and saying, “Make me a futuristic-looking spaceship.” Doing that would leave him without these insights, without that grounding, without knowing where it came from. ## From Blender to Godot: Understanding Comes from Doing It is like looking at C# code and suddenly thinking, “Wait, this function can do all that?” That means I can actually… He had no idea how well Blender and Godot worked together, and he had previously been exporting things from Blender in a really stupid way. Now he keeps everything’s parameters exposed. So when he makes something like the switch on the oscilloscope, making the red thing inside is actually difficult, and he probably would not do it because the movement is so precise. For the switch’s actual movement, he can access that pivot point from Blender inside Godot and use it to rotate the switch and drive the animation, instead of simply exporting the object. He says perhaps everyone already knows this, but it was precisely by doing the work himself that he came to understand and learn these things. It opened up his thinking, influencing and inspiring him to think of all the other things he can now do. Whenever he writes, he likes to keep a notepad open, filled with ideas that come from these experiences. ## Writing: Letting Characters Emerge Through Repeated Rewrites Will any of that get used? Maybe he will cut the bit about gloves. Maybe it will not work. Maybe it will come up somewhere else. He will keep writing and rewriting, producing a lot of material. Everything he does, every part he engages with, every old system and old technology he sees influences and inspires what goes into the game, what he writes, and who the characters are, because he is exploring who they are. There is no book telling him who everyone is. He knows how he wants the story to end and what state of mind he wants players to be in. He has a beginning and a lot of the middle, but he does not know everything. So by writing this much, he can find the things that really tell you about the characters, the things that reveal who they actually are. ## The Example of Ellis: Through the Writing, Not by Saying It He used Ellis from Left 4 Dead 2 as an example: he could have been written as a not-very-bright country boy, but almost everyone who has played understands that he is just playing dumb and being endearing, and is actually very smart. You accomplish that through the writing, rather than having him say it. Often, that happens through a dialogue with physical reality. So if he lets AI do these things directly, hand him a world, and then he enters it in a dry, detached way, he loses all of that: all the thinking, all the understanding of why these things are the way they are and how they connect. ## Slow, but Worth It There are all sorts of other reasons why his progress is slow, but he is willing to spend that time exploring. He got a new tooth, a dental implant, but his body is rejecting it and it hurts, so there are many other reasons why everything takes a long time. Still, that time is worth spending, especially early on—he is laying the foundation for the game and understanding the feeling, look, and tactile quality he really wants. Suddenly—“Oh my God, I have to load a tape because I need to switch communications modes.” He remembers getting tapes in the ’90s, loading them into the machine, making sure the channel was selected, and then having to go through that process. When you know the goal, know how to do it, and simply have to go do it, there is also the process of finding it and all those interactions. There is a whole bunch of other stuff he will discuss later, such as their crazy computer and that noun–verb system—it is both incredibly powerful and incredibly simple, with an unbelievable number of tricks to it. Every time he talks about it, he remembers another 500 things he learned, then flips through his notes and sees all those really cool things. Now he almost wants to record the whole thing again. ## Advice for Beginners So if you are starting something new, do not begin by asking ChatGPT to “make this.” Instead, learn about it, immerse yourself in that world, and try making some of those things yourself. He shared his button and knob models on Discord and received a bunch of feedback, taking them from looking ugly to looking great. Again, he is not trying to copy them exactly, but to capture that feeling and appearance—and those are things you can only get through interacting with and discovering these objects. ## “The Carpenter in the Woods of Maine” He says he is not worried about sales or anything else about this game. He is fortunate to be at this point in his life—“I’m just that carpenter in the woods of Maine, alone in my workshop, waking up every morning, having a cup of coffee, and hanging out with my cat” (who is very greedy about food). Eventually, he will make something and put it by the roadside, without caring whether it sells, because he will simply move on to making the next thing, because he enjoys the process. Source: A case against AI in game development from Chet Faliszek (Half-Life, Portal, Left 4 Dead, etc.) (https://www.youtube.com/watch?v=g9koXZ5JhDU)