Neural networks trained on large datasets are used by Pareto's AI to understand input photos or questions. The AI then creates new graphics based on the parameters and input given by mixing learning patterns and styles. This results in unique visuals.