by More than 100 years ago, Alexander Graham Bell asked National Geographic readers to do something bold and fresh – “discover a new science.” He pointed out that sciences based on sound and light measurements already existed. But there was no smell science. Bell asked his readers to “measure smell.”
Today, smartphones in most people’s pockets provide impressive built-in capabilities based on sound and light sciences: voice assistants, facial recognition and photo enhancement. The science of smell offers nothing comparable. But that story is changing, as advances in machine olfaction, also known as “digitized smell,” are answering Bell’s call to action.
Machine olfaction research is a major challenge due to the complexity of the human sense of smell. Although human vision depends mainly on receptor cells in the retina – rods and three types of cones – smell is detected through about 400 types of receptor cells in the nose.
Machine olfaction starts with sensors that detect and identify molecules in the air. These sensors serve the same purpose as the receptors in your nose.
But to be useful to humans, machine olfaction needs to go a step further. The system needs to know what a particular molecule or set of molecules looks like to a person. For this reason, machine olfaction requires machine learning.
Applying machine learning to smell
Machine learning, and in particular a type of machine learning called deep learning, is at the heart of significant advances such as voice assistants and facial recognition apps.
Machine learning is also key to digitizing odors because it can learn to map the molecular structure of an odor-causing compound to textual odor descriptors. The machine learning model learns the words that people usually use – for example, “sweet” and “dessert” – to describe their experience when they encounter specific odor-causing compounds, such as vanillin .
However, machine learning requires large datasets. There is a huge amount of audio, image and video content on the web that can be used to train artificial intelligence systems that recognize sounds and pictures. But machine olfaction has long faced a problem of data scarcity, in part because most people can’t verbally describe smells as easily and identifiably as they can describe sights and sounds. Without access to web-scale datasets, researchers have been unable to train truly powerful machine learning models.
However, things began to change in 2015 when researchers launched the DREAM Olfaction Prediction Challenge. The competition released data collected by Andreas Keller and Leslie Vosshall, biologists who study olfaction, and invited teams around the world to submit their machine learning models. The models had to predict odor labels such as “sweet,” “floral” or “fruity” for odor-causing compounds based on their molecular structure.
The best performing models were published in a paper in the journal Science in 2017. The winner was a classic machine learning technique called a random forest, which combines the output of multiple decision tree flowcharts.
I am a machine learning researcher with a long-standing interest in applying machine learning to chemistry and psychiatry. The DREAM challenge piqued my interest. I also felt a personal connection to olfaction. My family traces its roots to the small town of Kannauj in northern India, which is the perfume capital of India. In addition, my father is a chemist who has spent most of his life analyzing geological samples. Machine olfaction therefore presented an irresistible opportunity for perfumery, culture, chemistry and machine learning.
Advances in machine olfaction began to pick up steam after the DREAM challenge ended. During the COVID-19 pandemic, many cases of smell blindness, or anosmia, have been reported. The smell, which usually takes a back seat, rose in the public consciousness. In addition, a research project, the Pyrfume Project, has made more and more datasets publicly available.
Smell deeply
By 2019, the largest data sets had grown from less than 500 molecules in the DREAM challenge to around 5,000 molecules. The Google Research team led by Alexander Wiltschko was finally able to bring the deep learning revolution to light. Their model, based on a type of deep learning called graph neural networks, established state-of-the-art results in machine olfaction. Wiltschko is now the founder and CEO of Osmo, whose mission is to “bring smell to computers”.
Recently, Wiltschko and his team used a graph neural network to create a “primary odor map,” where similar odors are placed closer together than dissimilar ones. This was not easy: Small changes in molecular structure can lead to large changes in olfactory perception. Conversely, two molecules with very different molecular structures can smell almost the same.
Such progress in cracking the smell code is not only intellectually exciting but also has very promising applications, including personalized perfumes and fragrances, better insect repellents, new chemical sensors, disease detection soon, and more realistic augmented reality experiences. The future of machine olfaction looks bright. It also promises a good smell.
This article is republished from The Conversation, a non-profit, independent news organization that brings you reliable facts and analysis to help you make sense of our complex world. It was written by: Ambuj Tewari, University of Michigan
Read more:
Ambuj Tewari does not work for, consult with, own shares in or receive funding from any company or organization that would benefit from this article this, and has not disclosed any relevant connections beyond their academic appointment.
