“Evolutionarily, these [smell and taste] are the oldest and seemingly, most primitive senses. However, we have yet to formulate a unifying theory about how they work, much less reproduce them in high fidelity in a VR-like environment.”

— Ilya Kolb Ph.D.,Senior Project Manager

Our Senior Project Manager, Ilya Kolb, shares his expert commentary on our ability to differentiate smells in this first installment of our monthly thought leadership article series from members of the Aromyx team. We hope you enjoy his insights!

How many unique smells and tastes have you encountered today? How about in a week? A year? Your lifetime?

These questions may seem esoteric, but they are highly relevant to an emerging trend of quantifying sensory perceptions using machine learning and AI. The goal is to impart quantifiable metrics and mathematical relationships onto sensations, to accurately describe them, digitally transfer them to other people, and create entirely new experiences.

From a sensory perspective, the last 20 years have seen incredible advances in digitizing the sense of sight. Today’s Virtual Reality (VR) systems can persuade us that we are walking in a meadow, in orbit around Mars, or helplessly falling from a skyscraper. Behind these illusions is a  technological stack of light-emitting diodes, computer processors, numerical methods, and a deep understanding of human vision.

What about the world of smell and taste? Evolutionarily, these are the oldest and seemingly, most primitive senses. However, we have yet to formulate a unifying theory about how they work, much less reproduce them in high fidelity in a VR-like environment.

The questions presented earlier lie at the heart of the challenge in measuring smell and taste. To understand the sense of smell, we must first map out its extent. In 2014, scientists from Rockefeller University concluded, to much fanfare, that humans can distinguish more than 1 trillion unique odors (Bushdid et al., 2014). This would mean there are more unique odors than stars in the Milky Way (100-400 billion). To put it another way, if you could smell a unique odor every second of the day, it would take you >30,000 years to smell all of them without a single repetition.

While our intuitions cannot be trusted when dealing with such large numbers, most people would agree that 1 trillion unique odors seems high. The finding was soon challenged (to significantly less fanfare) by two groups who questioned the study’s experimental design and mathematical analysis. Both studies concluded that 1 trillion odors is an unlikely upper limit but the real number is likely several orders of magnitude lower (Gerkin and Castro 2015, Meister 2015).

The actual number of unique smells is still a mystery. Fewer distinctive smells for food, flavor, and fragrance industries can be considered good news: it gives hope that custom-formulated foods, beverages, and perfumes could accurately mimic natural flavors, offering us healthy alternatives to everyday products as well as new delights to enjoy. For scientists, fewer unique smells is also good news because it makes the problem of mapping out the olfactory space more manageable. However, complexity creeps back in when one considers the variability in perception between people—although two people may smell the same exact odor, their perception of it may be different.

We are at an exciting time in smell and taste research. We can leverage recent advances in biological assays, laboratory automation, and machine learning to tackle this complex perceptual space. The ability to map and use it can open the door to new frontiers in product development. On a more basic level, understanding smell and taste can help us communicate better with each other. Much like the radio, television, and most recently, VR, has completely transformed every aspect of our life. Unleashing the power of taste and smell could also revolutionize how we interact with the outside world.