For people, smell is part of life and romance. For example, it is always said; the smell of freshly baked bread, the smell of earth soaked by the rain, and the memories it brings back... In this way, we have learned and distinguished fifty thousand smells. If we wanted to address smell in technology, the question should certainly have been asked: “Can Artificial Intelligence be taught smell?”
Our nose; an engineering miracle our nose is a must-have organ of our body, much more than a simple structure of cartilage. Especially at this time when we are fighting strange diseases, I will first start with our unique characteristic reflection of each person, what benefits our nose has for us and why we want to be imitated.
His first task is breathing. The air taken from the nose is heated inside, moistened with mucous (slime) and filtered. Even our nose hair helps the filtering process. We all say “ew” because we portray them in our eyes when we read them for sure :)
Another task, which is our main topic, is to smell. In fact, it also helps to taste when smelling, which makes the food taste more delicious. That's why we say when we have the flu, “I didn't understand anything from what I ate.”
In fact, our nose has three other tasks that I can share information about later for those who are curious. Let's move on to “being able to smell”, which is our main topic …
I'll leave a fun video of how our nose can smell here. After watching it, I suggest you continue reading my article.
For people, smell is part of life and romance. For example, it is always said; the smell of freshly baked bread, the smell of earth soaked by the rain, and the memories it brings back... In this way, we have learned and distinguished fifty thousand smells. If we wanted to address smell in technology, the question should certainly have been asked: “Can Artificial Intelligence be taught smell?"Google's think tank is obsessed with this and is publishing an article in October 2019. “Using Deep Learning To Predict The Odor Properties Of Molecules”. According to this academic article, Google can teach machines how molecules smell using artificial intelligence based on their chemical structure.
How do they do it? Using graphical Neural Networks (GNN for short). GNN can be likened to algorithms that allow you to make friend suggestions when browsing social media today. Here, the structure of a scent molecule is translated into graphs, and with deep learning, these graphs are interpreted using GNN.
About five thousand fragrance molecules were distinguished by naming and taught to machines by fragrance and perfume experts, and artificial intelligence was expected to predict for molecules that were not named, and these tests were successfully completed. But there are some annoying details, for example, if a smell is “fruity” to me, it can be described as “floral” to someone else. There are molecules that have the same atom and bonds, while the form of sequence is opposite to each other. For example, fragrances such as vanilla and izvanillia, cumin and mint are different from each other and are called chiral pairs. The smell of vanilla smells like vanilla flower, while izvanilya feels like a bitter smell of Medicine. How artificial intelligence behaves in this situation is still a mystery.
In what areas will teaching artificial intelligence help us? For example, determination of freshness, such as whether food products are degraded, determination of space quality in hotels, cars or many places, detection of hazardous materials in industry. The companies Dai-ichi Seiko and Toppan Printing shook hands to work together to identify odors in the areas and produce a solution against the danger. Dai-ichi Seiko produces electronic equipment, while Toppan Printing Works on the artificial intelligence side of the business.
Teaching scent to artificial intelligence is also seen as an opportunity to produce new scents. In this context, IBM and Symirse have developed an artificial intelligence that can learn about top trends in the industry, successful data obtained in the past, and raw materials. IBM, which works with Symrise for three business lines, aims to achieve success; beauty, care and perfume culture. The goal is to discover innovations in the field of smell. For this purpose, IBM designed two different perfumes with artificial intelligence Philyra for O Boticário, one of the world-famous beauty companies.
It has also been developed to produce completely different scents with the artificial intelligence Philyra machine learning, which is expected to sound in the food industry, such as creating recipes, matching flavors.
By March 2020, when all these developments were happening, we read: "the chip that learns to smell has been developed!”. This development is announced from Intel. Using observations on mammals with a mathematical algorithm that would mimic the olfactory side of the brain, he developed norömorphic Loihi chips to identify and distinguish odors. These chips were taught to smell different odors and even dangerous chemicals by mimicking brain cell components. The chip can detect the smell of about ten dangerous chemicals, such as methane, acetone, ammonia.
According to some articles written on this subject again, scent detection of artificial intelligence, artificial intelligence, audio and video teaching efforts didn't much interest, but the “electronic nose system” or “imitation nose” in later times to facilitate the execution of jobs in many industries and contribute to human life looks like it will remain so.
Who knows, maybe one day it will help those who don't have a sense of smell, who can't smell fresh grass, newly brewed tea…