Artificial Intelligence can now comprehend visuals and sounds exceptionally well, but one sense that computers and AI have yet to grasp, at least for now, is the sense of smell. Osmo is a startup founded by Alex Wiltschko, who also serves as the CEO. Its goal is to develop AI technology that can generate scents based on conditions, similar to how AI creates images and text.
Wiltschko is an expert in olfactory neuroscience, having obtained his doctorate in this field from Harvard University and worked as a researcher at Google Research, focusing on developing Machine Learning capable of predicting scents from specified molecular structures. He later resigned to establish Osmo in 2022.
Creating desired scents through an understanding of molecular structures is Osmo’s short-term goal, while Wiltschko states the long-term goal is to develop smell-oriented AI to help people lead healthy and happy lives. The challenge lies in developing a computer that can perceive scents. He claims that smell is a factor that aids doctors in diagnosing diseases.
Developing smell-centric AI poses various challenges, from constantly changing adaptable molecular bonds to scents that can alter over time. Additionally, obtaining training data for AI proves difficult. While some companies have collaborated with certain fragrance manufacturers, they found the data sets unsuitable for training purposes.
Osmo’s data set utilizes Graph Neural Network (GNN) to analyze the connections between molecules in different relationship pathways, enhancing the understanding of scent patterns, resulting in better replicable scent production.
TLDR: Osmo, a startup led by Alex Wiltschko, aims to develop AI technology capable of generating scents based on specified conditions, challenging the development of computer systems that can perceive smells effectively.
Leave a Comment