A smart guy has already found a unique application for ChatGPT-4, the latest AI language model from the creators of ChatGPT. It turns out that GPT-4 can read the code for the Game Boy Advance game Pokemon Emerald and turn it into a fully interactive text adventure, with the mechanics of a Pokémon adventure.
Twitter user Dan Dangond discovered that the AI model could run Pokemon Emerald, a popular Game Boy Advance game released in 2005 in the US. Dangond shared a screenshot-by-screenshot thread of their conversation with the chatbot. The game starts more or less as it would if you were playing it on handheld – save for a few features, like setting the clock at the start – but it works like a fully interactive text adventure. Dangond “plays” the game by responding to the AI in the chat.
As always with regular Pokémon games, in Pokemon Emerald you explore a pixelated world and battle other Pokémon in turn-based battles, so the text largely describes battles and calculates them. As Dangond begins, the AI offers the choice between the region’s three starters: Treecko, Torchic, and Mudkip. From there, the Mudkip is sent out to battle a Poochyena, a real Pokémon you can battle on the first route.
Launches the game without any problem, fairly accurate so far, skipping the moving truck and clock setting, which frankly isn’t the most immersive thing about this game anyway pic.twitter.com/o4nqtMdCHH
— Dan Dangond (@DanDangond) March 15, 2023
The rest of the game runs like a raw text simulation that incorporates all of the main Pokémon mechanics, such as the type’s strengths and weaknesses. There are several missteps though, for example it can’t calculate that the insect/soil type Nincada is weak to water. It also randomly decided to include accuracy calculations after playing part of the game, and let a Rock Tomb attack miss. It also skips several battles and misses certain beats.
This text-based game comes at a time when many people are still sensing the true capabilities of generative AI. According to OpenAI, GPT-4 is a multimodal model, which is “less capable than humans in many real-world scenarios” but can “show human-level performance across a variety of professional and academic benchmarks.” According to other technology journalists who have covered the technology, it is supposedly more capable and advanced that its predecessor, ChatGPT, and it is acing standardized tests left and right. Now we can add running a Pokémon game to this list of skills.
ChatGPT-4 was not created with video game play in mind, and it shows. Anyone who wanted to play the game this way would greatly benefit from playing the original game. Without visuals and general knowledge of the world it would be almost impossible to play. Still, it’s a nice discovery that GPT-4 can run this game in any capacity, which is where the potential for fun and more crafting lies.