AI Model Epicure Maps Global Cooking in Two Megabytes
- Epicure, developed by KAIKAKU.AI, is trained on a dataset of 4.14 million recipes from seven languages.
- The model compresses culinary data into a two-megabyte file, representing cooking knowledge mathematically.
- Epicure consists of three variants—Cooc, Chem, and Core—that answer different culinary questions using the same data set.
- It maps out a total of 1,790 ingredients using a coordinate system based on their behavior in global dishes.
Epicure by KAIKAKU.AI compresses extensive culinary knowledge into a compact format, offering insights into ingredient pairings and substitutions without storing actual recipes. This innovative approach allows users to explore global cooking traditions through mathematical coordinates.
With its unique ability to map ingredients across various cuisines and contexts, Epicure provides reliable culinary insights while maintaining a small digital footprint. (Source)