Визначення перспектив застосування штучного інтелекту для генерації рецептур енергетичних батончиків

The objects of research are energy bars developed using mathematical modeling and various artificial intelligence (AI) models, including ChatGPT, Gemini, and Claude. The paper analyzes the quality and shelf-life indicators of these products. The AI application in recipe addresses the challenge of overcoming the complexity of traditional mathematical modeling for the rapid optimization of multi-component recipes. This ensures the creation of energy bars with an enhanced nutritional profile without compromising their sensory and structural-mechanical properties. The “Horikhovo-fruktovyi” sample consisted of oat flakes, dried cranberries, and prunes with a nut mix of almonds and peanuts, and chaenomeles. The composition of the “Shypshyna” bar was generated by the Claude AI model and contained whey protein and oat flakes, as well as rosehip, honey, and nuts. The “Askorbinka” sample, generated by Gemini, contained soy components: protein isolate, milk powder, and oat flakes. The recipe for the “Smorodyna” bar, generated by the ChatGPT model, included protein isolate, oat flour, and berry powders. The use of AI allowed for an improvement in the protein profile. The protein content in the “Askorbinka” and “Smorodyna” recipes is 2.3 times higher than the protein content in the “Horikhovo-fruktovyi” sample. These data may be explained by the fact that AI databases contain information indicating that energy bars should have high protein content. The organoleptic evaluation of the energy bars was carried out using a 25-point scale developed by the authors. Among the fresh products, the highest score (24) was achieved by the “Smorodyna” sample, the recipe for which was generated by ChatGPT. The samples were stored for 14 days in various packaging materials. The sample generated by Claude exhibited the best organoleptic characteristics. Regarding moisture content and acid value, the “Horikhovo-fruktovyi” sample performed best, showing moisture values from 19.9% to 25% and an acid value ranging from 1.75% to 1.83% at the end of the storage period. It was established that parchment paper and foil possess the best barrier properties for sample storage.