A Smarter Diet

Will the food of the future be based on AI-created recipes?

- Matthew Flynn

Anyone who has spent even the slightest amount of time on social media has witnessed the proliferation of pictures of food. Along with these shots of healthy salads, juicy steaks, and indulgent desserts come descriptions and reviews.

All the collected data is ideal for developing AI and machine learning, which can then use the information to create new recipes. Additionally, artificial intelligence can be used to create food that optimizes nutritional value and flavor by using the knowledge gained from millions of pictures and descriptions.

REVERSE-ENGINEERING FOOD

MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL), working with the Qatar Computing Research Institute and Universitat Politecnica de Catalunya, developed Pic2Recipe, an artificial intelligence system that discovers recipes from pictures of food.

The artificial neural network (ANN) read more than a million recipes and found patterns between ingredients and the pictures. Pic2Recipe was largely successful but needs some further refinement when evaluating pictures of food that contains many different ingredients, such as lasagna.

As the technology is refined, neural networks such as Pic2Recipe can help consumers know what is in their food and the exact amount of each ingredient. Such an advancement could be of great significance to those with allergies or those who are trying to adhere to very specific diets.

AI COOKING ASSISTANTS

Similar technology is currently available in the palm of your hand. Mobile apps such as PlantJammer and Chefling help create recipes using ingredients that are already in your house. PlantJammer generates vegetarian recipes based on users’ preferences by using existing aroma profiles to combine ingredients in different ways. The app also allows users to find ingredients to add based on desired flavor and texture profiles.