Can AI Smell Too? The Emergence of New Media Art with Olfactory AI

AI writing poetry or creating art is a familiar sight today. Beyond that, AI capable of identifying scents has emerged, causing astonishment. This development has caught the attention not only of the scientific community but also the art world, exploring possibilities such as using scents in new media art and digitally presenting scents in the metaverse.

AI-based olfactory technology startup OSMO conducted collaborative research with the US Monell Chemical Senses Center, Reading University, and Arizona State University, publishing a remarkable conclusion in the international journal ‘Science.’ OSMO, a startup derived from Google Research’s machine learning studies, aims to digitize scents by combining AI and olfaction science.

The ability to detect scents like the smell of baking bread, fresh grass after rain, tobacco smoke, or a grandmother’s perfume is based on the mechanism where odor molecules float in the air, enter our noses, and bind to olfactory receptors, transmitting information directly to the brain.

While the sense of smell offers various intriguing artistic aspects, the ‘Proust Effect,’ which vividly conveys associated memories and emotions, particularly enhances its appeal. Unlike other types of sensory information, smell takes a direct route, transmitting directly from sensory organs (the nose) to the brain’s centers of memory and emotion, making it the most instinctive and subconscious stimulus. This direct pathway is considered to hold immense potential, as it evokes specific and powerful memories.

However, the scientific classification of smell has been considered highly complex and challenging, leading to it being regarded as a rather subjective sense until now. For instance, while the human eye employs only three types of cells and around 40 receptors to detect colors for visual information generation, the nose boasts approximately 400 types of receptors or sensor proteins. Consequently, summarizing and classifying scents due to their intricacy has proven far more challenging compared to other sensory experiences. Despite this complexity, artists and scientists continue their endeavors to explore this captivating sense.

As part of this exploration, Wiltchko from OSMO has made strides by building a ‘smell map,’ an innovative method to classify scents, using artificial intelligence instead of relying on the human nose.

Their developed AI surpassed human participants in recognizing and distinguishing the scents of unidentified liquids contained in 400 bottles. It also succeeded in distinguishing structurally different pairs of molecules that emit similar scents, and even exhibited the remarkable ability to predict scents of substances it had not been trained on.

Alex Osmo, CEO of the company, stated, “This study marks the first step in quantifying smell. Computers have been able to digitize vision and hearing, but smell, one of the oldest and deepest senses, has remained undigitized. Smell plays a vital role in human survival, tightly intertwined with emotions and memories. However, we were unable to develop a systematic method for quantifying this sense. Now, we can control and engineer scents, opening the door to the next phase of olfactory innovation that can contribute to human health and well-being.”

This research process ultimately delves deep into understanding how the human brain interprets odors, captures scents, and retains olfactory memories, leading to the development of new techniques for manipulating these processes. Particularly in the realm of art, the utilization of such advanced technologies opens up the possibility of digitizing the sense of smell and creatively employing it within the metaverse. Leveraging fragrance in art combines intricate rationality with delicate sensibility, offering a visionary genre that innovatively merges reason and emotion.