Creating a Functional AI Chatbot in ‘Minecraft’

In Brief

A dedicated Minecraft player has meticulously constructed a functional GPT-style chatbot using the game’s redstone, avoiding any mods or command blocks.
The innovative model comprises approximately 5,087,280 parameters and is built over a massive space of around 439 million blocks, the result of several months of work.
Operating largely as a conceptual tool, the chatbot’s processing speed is notably slow, with response times that could extend up to nine years in real-time gameplay.

Redstone as a Computational Tool
The realm of Minecraft saw an impressive technological feat this past weekend as a GPT-style chatbot, wholly built from redstone, was unveiled. Redstone in Minecraft is used to simulate digital activities such as power flows and logic operations and is commonly harnessed to build a range of devices from simple mechanisms to complex computing systems.
Technical Details and Construction
Created by a user named Sammyuri, this AI machine was designed using a pre-trained language model based on the TinyChat dataset. The AI model within the game possesses about 5,087,280 parameters, a 240 embedding dimension, and operates with six layers within a 64 tokens context window. To conserve integrity while managing such a grand scale, the model primarily uses 8-bit quantization for weights, with some exceptions for embeddings and LayerNorm weights that require greater precision.
Physically, this extensive system covers an area totalling approximately 439 million blocks, in dimensions that stretch about 1,020 x 260 x 1,656 blocks. Sammyuri utilized the Distant Horizons mod to visualize these immense structures in the expansive setting of Minecraft. The intricate network of redstone circuits performs various complex operations essential to neural computation, including matrix multiplications and softmax approximations, similarly to how traditional computing operates.
Processing Speed and Utility
Despite the ingenuity and precision in its assembly, the chatbot functions at a speed far from practical. Even when the game’s tick rate is increased to about 40,000 times its normal speed, the system takes several hours to produce a single response. At regular speed, completing one operation could surpass nine years, indicating a substantial performance compromise for this experimental design.
Cultural and Educational Impact
Sammyuri’s work transcends typical gameplay, providing a vivid illustration of how standard computational theories can be adapted to unconventional mediums. This project not only alters perceptions of computational boundaries but also sparks broad discussions on potential future applications in fields like optics, DNA computing, or mechanical systems. Online communities have engaged in spirited discussions about the technical complexities and philosophical implications of such computational explorations.
Conclusion
While not a viable option for everyday computational tasks, the significance of this project lies in its ability to provoke discussion and foster innovation. As a pioneering technical artwork, it proves that computational technology can extend well beyond traditional environments. This model serves as an intriguing proof of concept and a subject of fascination, suggesting that the synergy between technology and creativity knows no bounds.