The electronic language made by AI is useful – Mael Business Newspaper

A research team at Pennsylvania State University in the United States announced that they had created an ‘electronic language’ based on artificial intelligence (AI). The results of the research were published in the academic journal ‘Nature’ on the 10th.

According to the researchers, this technology can quickly and accurately identify spoilage in milk with different water contents, as well as soda and coffee mixes, and fruit juices, as well as issues related to food safety. . The researchers predicted, “It could be useful not only in food production and safety, but also in medical diagnosis. “

The electronic tongue sensor is a graphene-based device that detects chemical ions and is connected to a neural network learned through AI. In an initial study, the research team proved that the electronic tongue can identify a variety of samples, including the freshness of milk, soda, coffee mix, and fruit juice, with more than 80% accuracy through 20 specified parameters with people. In subsequent research, AI applied a method to analyze data on its own and make decisions, and with this method, accuracy improved to over 95%.

“In the process of developing the electronic tongue, we wanted to go beyond just simulating the function of the tongue, but also simulating the brain’s ability to detect small differences in food recognition,” said Saptarshi Das, Penn State University professor and lead researcher. of the study. In fact, the human tongue interacts with various chemical substances as well as the five basic tastes of food (sweet, sour, sour, salty and umami) to distinguish between tastes, and the gustatory cortex of the brain explaining this complex information. The research team analyzed the data by designing a neural network algorithm that has a role similar to the human gustatory cortex.

In particular, this study focused on analyzing how AI makes decisions. “We used Shapley’s additive definitions, used in game theory, to understand what data the AI ​​saw as important in the decision-making process,” said co-author Andrew Farnon, doctoral researcher. Through this, the AI’s decision-making process became clearer, and it was revealed that the neural network reaches decisions by comprehensively considering data that is considered important as well as the parameters it has set. people.

This electronic language has great advantages, especially in terms of flexibility and economy. Despite the imperfection of the sensor, AI can process them in an integrated way to make correct decisions, so there is no need for perfect precision during the production process of the sensor. Thanks to these features, the research team expects that this technology can be widely used in various industrial fields.

Professor Das emphasized the practical potential of the electronic tongue, saying, “We found that nature can accept some imperfection, just as it is not perfect but still makes decisions powerful “