Let's take a look at how AI is changing food production

Giuseppe has a taste for great food but it’s not a food blogger, a chef or anything in between — Giuseppe is one of the technologies being used to change food production at food-tech company NotCo.

With the rise of artificial intelligence (AI), machines are being trained to analyse ingredients, make recipes and consider the taste, functionality and smell of food the way a chef might. At NotCo, this looks like a platform that learns and understands the molecular composition of animal-based products and replicates it using plant-based materials.

The brand’s first three products available in Australia — NotMilk, NotBurger and NotChicken Nuggets — have been created using AI.

Matias Muchnick, CEO and co-founder of NotCo, said: “Artificial intelligence enables us to unlock a universe of 300,000 plants (the western diet comprises around 200 plants) and look for combinations humans would never think of to replicate the exact taste, texture, smell and functionality of animal products — removing animals from production.”

Giuseppe works by generating different recipes equivalent to the product NotCo wants to recreate. It then looks for matches in flavour, texture, nutrition and functionality. The product is then pushed through to the company’s chefs for testing, and they give Giuseppe feedback on the product. The research chefs test over 100 recipes a month.

The algorithm is built to get smarter as it develops and establishes the underlying patterns between molecular components in food and the human perception of taste, texture, smell and colour.

NotCo Research Manager Rodrigo Contreras said, “We nourish the algorithm with different data obtained from laboratory assays (chemical, physical, spectroscopical, chromatographic analysis, among others) to understand attributes, composition, behaviour and other properties (characterisation).

“We use different plant sources of proteins (for concentrates, isolates and texturised proteins), principally from legume sources (Fabaceae) such as chickpea, pea, and other beans; and others from certain cereal (Poaceae) sources, such as rice and wheat. All the protein sources are properly labelled in the packaging.

“This isn’t lab food. It’s food production that blends data, science, computers and real-life taste testing, bringing together software engineers, analytical scientists and chefs.”

The production time quickens with each product as the algorithm is fed information. The first of its products, NotMayo (yet to be launched in Australia), took 18 months to produce, yet one of its latest products, NotChicken Nuggets, took only two months. However, it hasn’t always been smooth sailing during product development. For example, when the first NotMilk was produced, the colour was green as the algorithm had to ‘learn’ that certain ingredients, such as dill, can alter colour.

NotCo was founded in Chile six years ago and its investors include Jeff Bezos, Roger Federer and Lewis Hamilton. The company’s AI food production format joins a range of technologies innovating the food system globally.

Image caption: iStock.com/carlosgaw