Around the world, researchers have been making strides in digitizing and replicating images and sounds. However, when it comes to senses beyond sight and sound, specifically smell, the work has been ongoing. Researchers have now harnessed artificial intelligence (AI) to tackle this challenge, as discussed in an article from New Scientist. The possibilities that arise when we can download scents from dubious websites hint at a new frontier for humanity.
Idelfonso Nogueira, a chemical engineer and professor at the Norwegian University of Science and Technology (NTNU), has demonstrated that AI can replicate scents more rapidly and precisely than traditional methods. In a study published on February 19th, Nogueira and his team outlined the learning process that enables AI to reproduce the effects of various olfactory experiences.
Initially, NTNU scientists sought to define the profiles of two specific scents by rating them on scales commonly used to characterize odors such as alcohol, coumarin, musk, and more. Subsequently, an artificial neural network was trained to generate molecular combinations to recreate these fragrances.
Merging molecules alone is insufficient to perfectly replicate a scent, as our perception of the odor can be influenced by how these molecules interact with the air or skin. For instance, a freshly sprayed perfume fades gradually in stages: first the “top notes,” then the “heart notes,” and finally the “base notes,” after hours or even days. AI shows promise in replicating this scent evolution.
Envisioning a future where all existing perfumes could be economically reproduced with great precision through an AI-calculated formula is not far-fetched. This would involve expanding the database of molecules, incorporating more complex ones, and further refining the various parameters that define a scent.
The implications of this research extend beyond the realm of perfumery. AI’s ability to recreate scents will enable future generations to receive more sensory information about environments altered by climate change or ancient ritualistic practices like mummification in ancient Egypt.
Despite longstanding efforts to isolate the role of each molecule in a scent, the field often faces challenges. Subjectivity in scent interpretation and the dynamic, ever-changing nature of scents due to environmental factors present hurdles. AI could potentially overcome the latter challenge.
The next question is how this dematerialization of sensory information will be utilized. Could we foresee the emergence of “scent printers” allowing individuals to reproduce fragrances at home? However, if issues arise such as empty “lemon” or “pine” cartridges (which are costly to refill), such innovations may prove less appealing.
