Researchers at Penn State University have unveiled an innovative "electronic tongue" that leverages artificial intelligence to perform a range of tasks related to food safety and chemical detection in liquids. Detailed in a paper published on 9 October in the journal Nature, this advanced system can differentiate between various coffee blends, determine the freshness of fruit juice, and identify harmful chemicals in water.
The technological breakthrough lies in the combination of an ion-sensitive field-effect transistor and an artificial neural network. The ion-sensitive transistor operates by detecting chemical ions present in a liquid and converting this data into an electrical signal. This signal is then interpreted by a computer enhanced with artificial intelligence, mirroring the processes in the human brain that lead to taste perception.
"We’re trying to make an artificial tongue, but the process of how we experience different foods involves more than just the tongue," explained Saptarshi Das, a co-author of the study and engineer at Penn State University. He elaborated that the artificial system mimics the function of the tongue's taste receptors and the gustatory cortex, the part of the brain that processes taste.
The AI component of this system functions similarly to the gustatory cortex by analysing and interpreting signals from the sensor. Initially, the researchers fine-tuned the AI with a set of predefined parameters to evaluate acidity, achieving an accuracy of 91%. When the AI was given the freedom to establish its own parameters, this accuracy improved to over 95%.
In practical tests, the electronic tongue demonstrated its capabilities by accurately differentiating between similar drinks such as soft drinks or coffee blends. It also detected when milk was diluted, pinpointed spoilage in fruit juice, and identified per- and poly-fluoroalkyl substances (PFAS), which are harmful chemicals, in water.
The researchers employed a method known as Shapley Additive Explanations to discern which factors the neural network prioritised when reaching decisions. This approach aids in demystifying the decision-making processes of artificial intelligence, a significantly challenging area in current AI research endeavours.
Das noted that the neural network excelled at recognising subtle data characteristics that elude human definition and that the holistic consideration of sensor characteristics by the network helps in mitigating daily variations, enhancing the sensor's reliability.
The research underscores the potential for this technology to adapt and compensate for imperfections, much like natural processes do. "We figured out that we can live with imperfection," Das stated, drawing a parallel between the resilience of natural systems and their electronic tongue. Such robustness opens up avenues for the technology's application in diverse settings where precise chemical discernment is crucial.
Source: Noah Wire Services