The AI sneaks up on the nose. Microsoft has developed a nose capable of detecting, for example, when a bread dough is at the optimum moment to be baked or when the food is in bad condition. Will we see artificial noses in kitchens soon? We asked its inventor, Benjamin Cabé.
“A long time ago, in May 2020, like many other people, I spent many hours trying to perfect my bread recipe and even trying to determine when my sourdough was in ideal condition to bake the perfect baguettes,” says Cabé.
With that idea in mind, he decided to build a small machine that, with the help of a microcontroller, some gas sensors, and a fan that helps direct odors towards them, collects the necessary data to train an open source AI model with in order to recognize the gas molecules that indicate that the dough has fermented properly and is ready.
“The Artificial Nose uses a neural network to correlate concentrations of gases in the air into odor categories. When connected to an IoT platform, the nose can be used for different things: it can help design a real-time alert system to know when a food has gone bad”, they explain from Microsoft.
The sensors of the Artificial Nose can detect concentrations of carbon monoxide, different volatile organic particles, among others, which would make it capable of knowing, for example, if what is placed in front of it is a cup of coffee. But just smelling, this nose is not enough. Just like the human nose, it must be connected to a brain or computer where the data is stored. This is how a certain percentage of particles in the air can be related to a certain smell. All the smells perceived will help to train the nose, just as a sommelier does with his wine tastings.
Cabé, who has also designed a 3D nose in case the machine wants to be attached to it, shares all the instructions, the code and even how much the materials cost (only $76.90) so that everyone who wants and knows can build your artificial nose.
On the uses of the Artificial Nose, he says the following: "it can be used for a wide variety of tasks, from helping people suffering from anosmia to detecting the smell of burning food, of milk that has turned sour." or to control the cleaning of office buildings”.
The author of Aromas del Mundo (Debate, 2021), Harold McGee, believes that artificial noses are "in an early stage of development." For him, his limitation is the following: “develop sensors that cover the wide range of structures that we can smell, train the AI system to identify, not only individual structures, but also the mixtures that all odors contain, and then correlate them with human sensations.” McGee believes that the most immediate uses in time will be to detect problems: if food has been burned or overcooked, gas leaks or spoiled food. "But the human nose won't be replaced anytime soon!"
Asked about the Artificial Nose, François Chartier is clear: “for the moment, the human nose continues to be the reference. The human experience reaches a level of detail that includes each individual's own culture, as well as their innate sense, their learned knowledge and their own interpretation of their environment. However, he sees it as logical to couple an electronic nose to the AI to give the sensors of aromatic molecules a sharper sense. “But at the moment, the AI doesn't have the senses that humans have. That is to say, the emotions linked to our feelings, the pleasure linked to smell, even the affectivity generated by the memories that certain aromas provoke. Although, who knows what the near future holds for us?
Chartier believes that we may begin to see this type of technology implemented in fast food outlets and restaurants in health centers. "In the haute cuisine sector, it will take longer, since chefs and sommeliers already have organoleptic capabilities that are vastly more complex and complete than the electronic noses currently on the market, with or without AI."
For the chemical engineer and popularizer Claudi Mans, author of Química en la cocina (Tibidabo), the limitations of an AI nose are three: the availability and variety of sensors, the sensitivity of the sensors used and the availability of data from the scent library. Examining his applications, Mans concludes that the cook's nose is as sensitive as the apparatus for culinary applications. “I can't imagine this machine in home kitchens and neither in restaurants, where it would be difficult to identify something in an environment where twenty recipes are being prepared simultaneously. Yes, I clearly see its use to monitor office environments, detect 'sick' buildings and in terms of environmental work safety”.
