Oh dear. Goonswarm Games are shutting down after Running With Scissors cancelled POSTAL: Bullet Paradise due to the use of generative AI. It’s a bit of a saga this one.
I covered the initial announcement, along with a follow-up update in there where Running With Scissors attempted to defend the developer. The backlash only continued, and eventually RWS cancelled it as per the statement GamingOnLinux was sent on December 5th via Vince Desi, founder of Running With Scissors
Most of the time when people talk about plagarism in relation to AI it’s not actually plagiarism. Unless you are referring to people using image edit models to remix someone else’s work, but you could say the same about photoshop or making a collage. These mostly come from misunderstanding how the models work, and there is a reason you don’t see technical people or machine learning engineers arguing this.
I do agree though that there is an issue with people becoming mentally unstable as a result of using LLMs or VLMs. There is a specific model family that caused this primarily due to alignment issues. That model being GPT-4o. To some extent other models did also contribute to this, but the primary reason is GPT-4o. OpenAI and others have tried to fix this, but the community surrounding ChatGPT has been very resistant. To the point that GPT-4o was fully removed from ChatGPT but people demanded it be returned to them and unfortunately they got their wish. It seems people had become emotionally attached to the model. I think in this case the people and community surrounding the models are their own worst enemy. There are some interesting benchmarks on LessWrong by AI safety experts showing that some models are much better at detecting and handling psychosis than others. I believe Claude and Kimi models performed the best.
As for authenticity and intentionality: I think you might have a point for some use cases. It’s also important to bare in mind that image and video generation are only one tiny subset of AI and even they have some good uses. In particular they can be used to tell stories written and voiced by human beings. Here I am referring to things like Gossip Goblin, which use AI generated video, but all the stories being told are written by humans. The GenAI here is being used instead of manually doing animation and special effects. One of the biggest uses of AI is in programming. This is used in everything from the latest Windows and Linux OSes to video games and websites. I don’t really see how using AI for writing code can remove intentionality from the process of making a game or other interactive media experience.
Edit: also you might want to read this: https://www.lesswrong.com/posts/6ZnznCaTcbGYsCmqu/the-rise-of-parasitic-ai
Re your last point, I’m a full time web developer and while I’m not building entire games or whatever, I am currently working on a fairly complex and involved data migration project. My boss has demanded I do the whole thing with AI and don’t write code.
Thus far, it’s been incredibly frustrating to get it to do what I need it to do without having the chatbot change tonnes of tiny things or assume and hallucinate stuff that simply shouldn’t be there. Beyond those time-wasting frustrations, the fact that I’m not getting hands-on means my mental model of how the data is translating from one system to another is muddy. It’s not as clear as it would be were I building this thing myself. Specifically, because I’m not building it myself, I’m not running into edge cases personally and unpicking the knots of the current system.
There’s no intentionality in what chatbots generate, by definition. They have no intention, they’re not alive, they can’t think. They don’t understand things.
I’m sorry but I’m sort of done with this topic. I don’t like Generative AI, I think it’s disingenuous, lazy, furthering the commodification of art and creativity, and damaging our abilities to think critically. However, I do understand that some people have found it helpful in some contexts, and other people like to play with it. Thanks for the chat. 👍