Since the early 1900s, the entertainment industry has been trying to pair the sense of smell with video entertainment.
In 1916, the Rivoli Theater in New York City introduced a fragrance to the theater during a movie called. The Story of Flowers. In 1933, the Rialto Theater installed a theater ventilation system. Hans Laube invented a technique called Scentovision, which was presented at the 1939 World’s Fair. Ten years ago, Japanese researchers were also experimenting with “Smell-O-Vision” for home TVs, working on a television that uses gel pellets that smell and release air streams from each corner of the screen into the living room.
However, none of these efforts took off, mainly because they didn’t work very well. These attempts at Smell-O-Vision have failed because we have never been able to recreate the world’s smells accurately or measurably, largely because we have never been able to capture them digitally.
This is not to say that the fragrance and fragrance industry has not strengthened and grown, but it is a very different task to create a unique scent for a consumer product than to build something like a “scent printer” that emits scent on command. The latter requires a complete digital understanding of odor molecules, something that has only just happened.
The digital understanding of the fragrance world has grown rapidly in recent years, and one company leading the way is Osmo, a startup that raised $60 million in funding. Osmo is led by Alex Wiltschko, a Harvard-educated, former Googler who received a PhD in olfactory neuroscience from Harvard in 2016. Wiltschko, who led a team at Google that spent five years using machine learning to predict how different molecules will smell, founded Osmo in early 2023 with the goal of improving “digital health” and improving human health. basic skills to make computers do our best. “
Osmo used AI to explore the connection between molecular structure and odor perception, showing that the machine can predict odors with incredible accuracy. They built a machine learning model using graph neural networks (GNNs), trained on a dataset of 5,000 known compounds, each labeled with descriptive odors such as “fruit” or “floral.” The model was then tested on 400 novel computers, chosen to be structurally different from anything previously studied or used in the perfume industry, to see how well it could predict their scents compared to human panelists.
The model’s abilities were also challenged in the “opposite” test, where it had to predict the odors of molecules that were structurally similar but had different odors. Osmo’s model correctly predicted odors 50% of the time in this difficult situation. Additionally, the model was able to generalize beyond the original training data, testing other olfactory properties such as odor strength on a large dataset of about 500,000 odor molecules.
The Primary Odor Map (POM) created by Osmo’s model outperformed human participants in predicting the harmonious odor of molecules, marking a major advance in the science of smell and showing that AI can predict odors based on molecular structure better than individual experts in many cases.
We have been able to digitally capture and categorize other categories of sensors, such as vision, which has led to the creation of a large number of new industries in robotics and autonomous vehicles. Big leaps have been the result of machine learning models, and now we’re seeing another big leap forward in power and product innovation through the use of generative AI.
Another potential application that Wiltschko describes is “teleporting scent,” where we would be able to capture a scent from one part of the world and transmit it to another. To do this, he envisions a world where a cell’s AI-guided spatial sensor can instantly identify the molecular structure of any odor. From there, his scent map can create what is a formula ready to be teleported without significant intervention by scent experts.
This idea, using AI to quickly recreate smells based on a digital frame, could lay the foundation for what film and TV makers have been dreaming of for a long time: creating technology that can recreate scents and smells at a high level. In other words, we can finally enter a world where Smell-O-Vision becomes a reality. The power of video entertainment, virtual reality, and other experiences in food service, tourism, and more can undoubtedly lead to a host of new applications, just as we’ve seen over the past few decades with advances in computer and machine vision.