AI is solving a hard problem: Giving computers their sense of smell

Over 100 years ago, Alexander Graham Bell asked National Geographic readers to do something bold and original.Founding a new scienceBell pointed out that a science already existed based on the measurement of sound and light, but a science of smell did not. Bell asked his readers to “measure smells.”

Today, the smartphone in most people’s pockets is packed with impressive features based on the science of sound and light: voice assistants, facial recognition, photo enhancements, and more. The science of smell has no parallel. But that’s about to change, as advances in machine olfaction (also known as “digital smell”) are finally answering Bell’s call to action.

Research into machine olfaction faces major challenges due to the complexity of the human sense of smell, which is primarily based on human vision. Receptor cells of the retinaThe olfactory system consists of rods and three types of cones, with approximately 400 types of Nasal receptor cells.

Machine olfaction begins with sensors that detect and identify molecules in the air. These sensors function similarly to receptors in the human 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 smells like to a human. To do that, machine olfaction needs machine learning.

Applying machine learning to smell

Machine learning, specifically a type of machine learning Deep Learningis at the heart of incredible advancements like voice assistants and facial recognition apps.

Machine learning is also key to digitizing smells, as it can learn how to map the molecular structures of odor-causing compounds to textual odor descriptors: Machine learning models learn words that humans often use to describe what they experience when they encounter a particular odor-causing compound, such as “sweet” or “dessert.” Vanillin.

But machine learning requires large datasets. The web has an unimaginably huge amount of audio, image, and video content that can be used to train artificial intelligence systems to recognize sounds and images. But machine olfaction has long faced a data shortage problem, in part because most people cannot describe smells in words as easily and recognizably as they can describe sight or hearing. Without access to web-scale datasets, researchers have been unable to train truly powerful machine learning models.

However, in 2015, researchers DREAM Olfactory Prediction ChallengeThe contest is Data collected by Andreas Keller and Leslie VoshallA group of olfactory biologists invited teams from around the world to submit machine learning models that had to predict odor labels, such as “sweet,” “floral,” or “fruity,” for odor-causing compounds based on their molecular structure.

The best performing model is Journal Articles Science In 2017, a classic machine learning technique Random ForestThe winner was , which combined the outputs of multiple decision tree flowcharts.

I Machine Learning Researcher I have long been interested in the applications of machine learning to chemistry and psychiatry. The DREAM Challenge piqued my interest. I also felt a personal connection to the sense of smell. My family roots are in Kannauj, a small town in North India. Perfume Capital of IndiaAdditionally, my father is a chemist who has spent most of his career analyzing geological samples, so machine olfaction offered a fascinating opportunity at the intersection of perfume, culture, chemistry and machine learning.

Advances in machine olfaction have accelerated since the end of the DREAM Challenge. During the COVID-19 pandemic, many Smell disorder or anosmiaThe usually-neglected sense of smell has been elevated in people’s consciousness. Pilfume ProjectWe have made many more and larger datasets publicly available.

Sniff deeply

By 2019, the largest dataset from the DREAM Challenge had grown from fewer than 500 molecules to approximately 5,000 molecules. Alexander Wiltschko They have finally succeeded in bringing the deep learning revolution to machine olfaction. Their model is a type of deep learning Graph Neural NetworksEstablished Cutting-edge results Wiltschko, who is currently active in the field of machine olfaction, OsmoOur mission is to “give computers a sense of smell.”

Recently, Wiltschko and his team used graph neural networks toMain odor mapIn olfactory matching, perceptually similar odors are placed closer to each other than dissimilar odors. This wasn’t an easy feat, since small changes in molecular structure can lead to big changes in olfactory perception. Conversely, two molecules with very different molecular structures can smell almost the same.

These advances in cracking the code of smell are not only intellectually stimulating, but also have very promising applications, such as personalized perfumes and fragrances, better insect repellents, new chemical sensors, earlier disease detection, and more realistic augmented reality experiences. The future of machine olfaction looks bright, and we can look forward to smelling good again.

This article is reprinted from conversation Published under a Creative Commons license. Original Article.