AI could see and hear. Now, it can smell!

Scientists at Google are busy using deep learning to make AI capable of predicting “the Olfactory Properties of Molecules”

Relationship between a molecule’s structure and its scent is difficult to predict, according to researchers. In order to find solution to this issue, Quantitative structure-odour relationship (QSOR) models are used. QSOR modelling is one of the unconventional Quantity structure-activity relationship (QSAR) models. Google uses graph neural networks (GNNs)to directly predict the odour descriptors for individual molecules, without using any handcrafted rules”.

If AI systems are able to predict smell, it will be very useful in streamlining human memory. Also, intellectual property system will become more certain about protection of smell as copyright. In the EU, there are two views about protection of scent as copyright. The Netherlands  recognizes it, and France doesn’t.With this technological advancement, smell could be digitized and, thus, could be protected under copyright law. Also, according to Google, it will be possible to develop and manufacture new smell molecules at very competitive price.

It seems that scientists are getting closer to affording, artificially though, senses to machines and systems that are not biologically humans.Could taste be the next on this list? At IBM, they think so.

Paramjeet Singh Berwal

Paramjeet, a lawyer, an invited lecturer, and AI research and policy consultant, is the Director of the Institute for Artificial Intelligence and Technology Law at Sulkhan-Saba Orbeliani University, Georgia. He is a Global Panelist with MIT Technology Review. He is frequently invited to deliver talks and presentations on various topics pertaining to AI. His research includes how AI will influence human existence, especially in the context of economy, work, law, society and its institutions, business management, social behaviour and policy making. He may be contacted at

Leave a Reply