Researchers in the United States and China have built an AI-assisted system that detects a smell in one place and recreates it in another, bringing the long-promised idea of digital scent a step closer to practical use. The prototype, called Scentronic, can currently handle five aromas: coffee, cake, orange, beer and red wine.
The work was led by a team at Ohio State University with collaborators at Zhejiang Laboratory in China, and was published on October 21 in the journal Science Advances. The lead author is Chunyu Yang, a doctoral student in materials science and engineering at Ohio State. The findings have since been widely covered, including in a report attributed to Reuters.
How it works
Scentronic has two halves. The sending side is an electronic nose, a sensor array that picks up the chemical signature of an aroma. According to Digital Trends, it uses six gas sensors. A machine-learning model then identifies which smell it is. That identification, not the smell itself, is what travels over the internet.
The receiving side is a compact wearable holding small reservoirs of scented liquid. When the signal arrives, tiny vibrating atomizers push the liquid through fine holes and release a mist that approximates the original aroma. The team built several versions, including a wristband, a handheld unit and a necklace.
No odor molecules cross the network. The receiver simply recreates an approximation from ingredients it already carries. That design is also the system’s central limit: every new smell needs training data for the electronic nose and a matching liquid loaded into the receiver.
How well does it work?
In tests with 14 volunteers, participants correctly identified the recreated smells about 73 percent of the time, Digital Trends reported from the paper. That is an encouraging result for a first-generation device, but it also means roughly one in four recreated aromas was not recognized.
Outside experts are interested but measured. Marianna Obrist of University College London, quoted by New Scientist, called Scentronic “a great idea” and a step in the right direction. She cautioned, however, that recognizing and releasing smells from a small, fixed library is very different from transmitting sound and vision.
Yang himself has been candid about where things stand. He told Reuters that the system is, in his words, a research prototype.
What it could be used for
The researchers point to three broad areas. The first is communication: sharing the smell of a meal with relatives abroad, or adding scent to a video call. The second is commerce, such as sampling a perfume before buying it online. The third is health. According to New Scientist, doctors might one day use such a system to assess distorted or lost sense of smell in patients with post-viral illness or neurodegenerative disease.
The team also built a prototype cup that pairs scent with flavored liquids, so a user wearing a mixed-reality headset could select a virtual drink and experience its aroma and taste.
What sets Scentronic apart from some rivals is size. Yang told CNET that Osmo, another AI-and-scent company, produces detailed chemical information but relies on much larger equipment, while Scentronic is designed to be worn.
Why readers across Africa should watch this space
Scentronic is a laboratory prototype, and nothing in the coverage reviewed for this report gives a price, a manufacturing plan or a release date. It would be premature to speak of African deployment.
Still, the project is a useful marker of where applied AI is heading. Much of the public conversation focuses on text, images and code. Scentronic shows researchers extending machine learning to sensory data, in areas, such as online retail and remote healthcare, where many African markets are already building digital services. The practical questions of cost, bandwidth and device availability will decide whether any such technology becomes relevant on the continent, and the research does not yet answer them.
A long road for digital scent
The idea of sending smell through media is not new, and its history is cautionary. The 1960 film Scent of Mystery released odors into cinemas, but the approach never took off, according to the Christian Science Monitor. Decades later, DigiScents, a company that set out to bring smell to the internet, shut down in 2001 after investing about $20 million, according to published biographical records of its former research executive.
Scentronic’s advantage over those efforts is that it is automated, wearable and driven by AI recognition rather than manual triggers. Its disadvantage is the same one that has dogged the field for decades: smell is chemically complex, and a library of five is a very small start.
For now, the honest summary is the one the lead researcher gave. This is a promising prototype, not a product.
Sources: Science Advances; Digital Trends; New Scientist and CNET (via secondary summaries); Reuters (via secondary reports); Christian Science Monitor.
