Artificial intelligence has made remarkable strides in recent years, enabling machines to see and hear. Now, Alex Wiltschko, a former Google employee, is taking these developments a step further by attempting to give machines the ability to smell. After spending five years on a Google team exploring the use of machine learning to equip machines with the sense of smell, Wiltschko founded his own startup, Osmo, in 2022. The goal of this company is to digitize the sense of smell, similar to how AI has successfully replicated human vision and hearing.
The Potential of a Robotic Sense of Smell
According to Wiltschko, digitizing smell could have revolutionary applications across various fields. In his initial announcement about the creation of Osmo, he highlighted that this capability could enable earlier detection of diseases, faster pandemic monitoring, improved food production and preservation, identification of spoilage before it becomes a problem, and even more effective insect repellent. Although these goals seem ambitious, Wiltschko believes his startup could make a significant impact in the short term.
Initially, Osmo will focus on creating sustainable aroma molecules, which could transform the production of perfumes, shampoos, detergents, and other similar products. The idea is to make these processes more efficient and environmentally friendly by using safer ingredients that are non-toxic and non-irritating to the skin or eyes. Wiltschko believes that his technology could enhance the quality of fragrances and other aromatic products, potentially posing a threat to the perfume industry, which could face a future of easily cloned fragrances thanks to AI.
The Challenge of Creating a Smell Database
During his time at Google Research, Wiltschko and his team used machine learning to create a "map of primary odors." This project involved analyzing 5,000 molecules responsible for different aromas, classifying them into categories such as fruity, floral, or mentholated. However, unlike images and text, there was no extensive database to train AI on smell recognition, so the team had to build a dataset from scratch.
Osmo's future lies in the ability to clone scents, a process that begins with digitizing smells and then replicating them using AI models. While this could represent a significant advance in fragrance technology, it remains to be seen how the perfume industry will respond to this innovation. Meanwhile, other developments, like Xiaomi's CyberDog, which can already detect certain smells, suggest that we are only at the beginning of what could be a new era in the world of artificial intelligence and the senses.