Big Walk and Little AI
If you're unaware of the controversy surrounding Big Walk, right now they've sparked off a discussion on LLM usage in games. Some background: They had disclosed to the IGF (The Independent Game Festival) that LLMs were used during the development process for bug fixing and 'consulting', which caused some ripples for the judges as many of them believe any use of AI is disqualifying. Earlier this week, this information broke out into the wider public and has sparked some intense debate and (intensely) disappointing takes from people I'd normally respect.
So I wanted to get out my thoughts here properly. I had been making smaller posts about it, but I wanted to really dig into my own feelings around this mess and why it feels so bad to me in particular.
I say this often, but I'm a software engineer. Problem solving is something I deeply enjoy, the process of identifying an issue, identifying the fix and surgically applying it to ensure it doesn't cause any downstream effects. Earlier this year, playing Big Walk resonated with me quite a bit and helped pull me a little bit out of the pit I'm in. The job market is hell and I've had no luck finding work despite my depth of experience, so Big Walk was a Little Oasis of getting to ignore that for a bit.
And the thing about Big Walk that I really enjoyed was that it mimics the process of software development. The game presents you a bunch of little puzzles, problems to solve with your friends. Everything is about communication, communicating concepts or ideas or shapes or sounds and putting it together for a solution. In my day to day life this is exactly what I did and what I prided myself on doing, we as software engineers aren't just folks that keep our heads down on a keyboard solo-coding in a dark cave until we solve the problem. I mean you can, but in my experience that was a very bad way to try and solve problems early on in my career.
So it was a game that celebrated the process of solving problems. The LLM usage in turn feels like a deep contradiction because it goes against that core ethos. It's easy and alluring to delegate your thinking to the LLM, to let it solve a hard problem for you. And according to them the Unity engineers they were working with said the problem they were trying to solve was impossible. I know from my own professional experience working with Unity engineers that this is a sort of thought-terminating response, a way to get you off their case even if the problem is very much so solvable.
It's not a huge use of AI, but to me it still feels like a mistake. But what makes me more disappointed than Big Walk's use of AI is how many people have simply shrugged at it. I don't think House House needs to be turbo-cancelled or anything for this mistake, but I don't think it needs to be justified either. And I see artists, people who would argue that everyone on a dev team matters, people who should really know better say that it's okay. That it was inevitable, that coding is just pure mechanical function and that what can you expect? There was no other way to solve this problem.
Reading those kinds of responses feels worse, while Big Walk's use of AI is a contradiction I'm disappointed with this feels more deeply like a betrayal. That we've given up on coding as an art form and have decided that it's okay to have a little AI here, as a treat. But if we don't have solidarity how can we expect to stand together when it creeps into things like art or music? They'll argue that there was simply no other way to create these textures and that they're a small part of the game, or that making specific sounds was too onerous for their studio. People have seemingly already forgotten that there was a time before LLMs, where problems were solved with coffee and manpages, digging through white papers, sharing insights with communities. People have decided all of that is too much so it's okay to use LLMs to help solve problems for your games.
I don't know. During my professional time at Blizzard I never really thought to consult LLMs when trying to solve problems because...why would I? I wanted to understand the codebase, why things failed the way they did so that I could build up that mental model of how things functioned. I wanted to reach out to my coworkers about specific bits they understood so that we could come to an understanding about the solution. That was at least part of the art of the profession. That's why I feel so baffled about the whole thing and the way people justify it. If only the solution matters, we're just opening a backdoor for LLMs to subsume all of art.
I hope I'm wrong. Maybe code will be the only thing sacrificed on the altar of LLMs and other professions will skirt by when the bubble hopefully pops. But I still love coding and will still continue to fight as much as I'm able to.