The software developer behind an unofficial PC version of N64 platformer Banjo-Kazooie has criticized the notion that generative AI tools expedite the creation of PC ports for classic games. DarioSamo, a member of the modder group responsible for the Banjo: Recompiled project, contends that the assertion that generative AI speeds up porting is either factually incorrect or fundamentally misunderstands the process.
He explained that decompiling a game, which involves converting its original machine code into human-readable code, should be an endeavor focused on comprehending and preserving the game's underlying design and construction. This process, he argues, is essential for creating robust ports for modern platforms.
DarioSamo recently addressed a video showcasing a Banjo Tooie "recompilation" project that utilized generative AI. He clarified that while the project used tooling developed for Banjo: Recompiled, it was not created by his team. He expressed concern that the video did not sufficiently distinguish this, deeming it "irresponsible" and potentially encouraging further AI-driven projects for fleeting recognition.
In a detailed thread on Bluesky, DarioSamo elaborated on the implications of using generative AI for creating PC versions of older games. He defined "recompilation" as a porting method using static recompilation to automatically convert a game's code for a new platform. These efforts often rely on decompilation projects for information to adapt the game for modern systems, including enhancements for high refresh rates, widescreen support, and quality-of-life improvements.
He asserted that claims of AI significantly accelerating this process are false, as recompilations were already achievable within a day by skilled developers even before the advent of current AI tools. DarioSamo highlighted a pattern where individuals believe the porting process is complete once the game is running, overlooking the extensive work required for polish and sustainability. "The pattern we've been seeing for several months so far is that a new type of person has entered the scene that thinks that because AI gets them over the hurdle of getting the game running, the job is basically done," he wrote. "And they need nothing but a credit card and a subscription to an AI service. What these people are unaware of is that getting the game running is basically the first step of the journey and not the end."
"This is why I argue that as far as recomps go, AI hasn't provided any sort of speedup," DarioSamo continued. "Making a barebones recompilation was always easy and required very little time. It's the polish and making a good and sustainable project for the wider public where the difference is. It's the part that actually takes far more time."
To illustrate his point, DarioSamo referenced the DK64 Recompiled project, a port of Donkey Kong 64 that does not employ generative AI. He noted that the Recompiled team was competing with an AI-assisted project. Due to their deep familiarity with the game, they achieved a superior state for their port, with DarioSamo providing additional support. In contrast, the AI effort reportedly stalled due to a lack of specialized knowledge.
While acknowledging that generative AI can expedite the initial decompilation phase to produce functional code matching the original ROM, DarioSamo stressed that the core purpose of decompilation is documenting and understanding the game's code. This is why decompilation and porting are often handled by separate teams: one focuses on documentation, the other on utilizing that documentation for porting.
He posited that Large Language Models often bypass this in favor of a brute-force approach, resulting in code that compiles correctly but is not human-readable or understandable, hindering subsequent port development and patching. Furthermore, using these tools means circumventing collaboration with other port developers and players, leading to a lack of awareness regarding critical development decisions and potentially resulting in unrecoverable projects.
DarioSamo concluded that the rapid proliferation of AI-generated ports can demoralize developers who adhere to high quality standards. He emphasized the importance of supporting these dedicated developers and discouraged the promotion of AI-generated projects in shared spaces, underscoring that an audience valuing meticulous work is crucial.
"These are old and beloved games that have been available in numerous ways for decades, so there's no rush to be the first to provide a solution using these tools, and it's unlikely it'll be the only solution in the future either," he stated.
The discussion also touches upon broader concerns within the creative industries regarding knowledge preservation, especially in the face of talent attrition and documentation loss. The embrace of AI for expediency, without regard for deeper understanding, mirrors a wider trend.
Similar sentiments have surfaced in other modding communities. The creator of a Total War: Warhammer 3 mod deliberately corrupted their work to prevent its use for AI-related purposes. Previous discussions have also explored the acceptable use of generative AI in modding and the need for platform transparency.