Depositphotos
Machines have long excelled at replicating sight and sound, but taste has remained much harder to capture in digital form. While previous efforts have produced artificial tongues specialized for sweetness, wine, or whisky, researchers in Beijing have now created a more versatile graphene oxide-based “tongue” that not only detects chemicals but also learns to recognize them. In laboratory experiments, the device identified sour, salty, bitter, and sweet with nearly 99% accuracy, proving that taste can indeed be digitized.
A Neuromorphic Artificial Tongue
Scientists at the National Center for Nanoscience and Technology in Beijing, working with collaborators across China, developed a neuromorphic device that mimics human taste. This “artificial gustatory system” uses layered graphene oxide membranes that sense and process chemical signals directly, much like natural taste buds working together with neurons.
Unlike vision or hearing technologies, which rely on solid-state electronics, taste must function in liquid environments where ions, not electrons, transmit signals. To meet this challenge, the team built a graphene oxide ionic sensory memristive device (GO-ISMD).
Inside its nanostructured channels, ions undergo adsorption and desorption processes that slow their movement, producing memory-like electrical responses. This short-term volatile memory enables the device to both sense chemicals and perform computations in liquid environments—a first in biomimetic systems.
Synapse-Like Responses
When exposed to voltage pulses, the device behaves like a biological synapse, adjusting the strength of its responses, retaining memories of prior signals, and showing extended recall times up to 140 seconds. To transform this behavior into perception, the researchers applied a computational approach known as reservoir computing.
“Our design was inspired by the biological taste system,” said Yon Yang in an email to New Atlas. “We built a smart platform with three main parts: a sensing input, a reservoir layer, and a simple neural network. The sensing input and reservoir are based on our hardware, which converts chemical signals into unique patterns that the neural network can learn to recognize.”
In practice, the sensing layer first captures flavor signals and translates them into electrical impulses. The neural network then learns these digital patterns, storing key parameters so the system can recall and identify flavors later.
Proof-of-Concept Testing
For their proof-of-concept, the team tested four tastants: sour (acetic acid), salty (NaCl), bitter (MgSO₄), and sweet (lead acetate). The system reached 98.5% accuracy in distinguishing them, with individual tests ranging between 75% and 90%. It also successfully classified beverages like coffee, Coca-Cola, and their mixtures.
Still, the researchers stress that this is an early demonstration. The current device is large and consumes significant energy, meaning further miniaturization and circuit integration are needed before practical use beyond the laboratory.
“This technology bridges brain-inspired computing, chemical sensing, and biologically inspired systems,” Yan explained. “With advances in scalability, energy efficiency, and multi-sensor integration, we expect breakthroughs in healthcare, robotics, and environmental monitoring within the next decade.”
By uniting sensing and computing in a single aqueous device, this graphene oxide system represents a major step in artificial taste and neuromorphic design—and it hints at future technologies that could expand, or even recreate, the human sense of taste.
Read the original article on: New Atlas
Read more: Robot Remorse: New study Aids Robots in Making Safer Choices Near People
