AI Taste-Smell Fusion Perception System GustoNet Deep Dive: Teaching Machines to Taste Freshness and Smell Purity

AI Taste-Smell Fusion Perception System GustoNet Deep Dive: Teaching Machines to Taste Freshness and Smell Purity

The human taste and smell systems are among nature's most sophisticated chemical analyzers. An experienced sommelier can distinguish over 800 volatile compounds in a single glass of wine. A pharmaceutical quality inspector can detect impurities at parts-per-million levels through smell alone. Over the past three years, the breakthrough progress AI has made in vision and hearing is now extending into the domain of chemical perception.

GustoNet was launched by SensAI, a startup spun out of ETH Zurich, in late 2028. The system employs a technique called "cross-modal perceptual fusion": raw signals from a 32-channel graphene-based gas sensor array (smell) and a 16-channel ion-selective electrode array (taste) are fed into a specially trained multimodal Transformer model.

SensAI Chief Scientific Officer Marco Bianchi explained: "The human tongue has roughly 10,000 taste buds, each containing 50 to 100 taste receptor cells. Our sensor array works on a similar principle — it doesn't try to identify individual compounds but instead constructs high-dimensional representations of taste and smell through the response patterns of multiple sensors."

In the food industry, GustoNet has been deployed in over 200 food processing plants across Europe and Asia. Japan's largest seafood processing company, Maruha Nichiro, reduced fish freshness detection time from 4 hours (including microbial culture) to 3 minutes after introducing the system in January. By analyzing trimethylamine and dimethylamine ratios in fish volatiles alongside pH and ionic strength signals, the system achieved a freshness grading accuracy of 96%, up from 78% with manual inspection.

In pharmaceutical testing, GustoNet has shown even greater potential. Indian pharmaceutical giant Cipla deployed a customized version at its Mumbai factory for detecting trace impurities in raw materials. The system can identify solvent residues below 10ppm — comparable to gas chromatography-mass spectrometry (GC-MS) — but 50 times faster, with results in just 90 seconds.

ETH Zurich sensory science laboratory director Professor Anna Keller expressed cautious optimism: "GustoNet's core innovation lies in combining the hard data of sensor signals with the soft knowledge of taste-smell fusion from neuroscience. But the subjectivity of chemical perception is a fundamental challenge."

SensAI says the next generation of GustoNet will add a tactile modality through microfluidic sensors that analyze food texture, achieving four-modal fusion perception.