The AI COOKBOOK challenges the seriousness that surrounds the topic of Artificial Intelligence. AI often associates with topics like Lethal Autonomous Weapons (LAWs) or facial recognition on social media sites. This project approaches the topic of machine learning through a humoristic and more human approach. Food is a relatively unexplored topic for machines, and very relatable for humans.
The project aims to both reflect our own obsession with digitising our edible fuel and the (current) incapability of machines to truly understand the physical world.
Based on large datasets, machines can identify flowers in photos and generate new images of cats, yet does this mean machines understand what a cat or a dandelion really are? If an AI can understand how a recipe is formed, does it understand how the ingredients work together? Or is it simply using the given building blocks to reconstruct what we expect from it, without considering things like flavour and taste.
This projects shows the disconnect between AI and the physical world.
It also provides social commentary on the notice that neural networks are somehow all knowing and all observant, which is sometimes the perception the general public carries. Whilst in reality most neural networks are highly specialised, and only work properly if fed the correct and a vast amount data. These datasets can come with a certain amount of biast depending on size and input. This is reflected in the generated images that are supposedly representative of the recipes. Whilst the generative adversarial networks network is not trained on food and thus unable to generate an accurate visual representation of the recipes.