"Electronic Tongue" Poised to be the First Step in Artificial Intelligence

Can AI get hungry? Can it develop preferences for certain foods? A research team at Pennsylvania State University is developing a new electronic tongue that can simulate how taste influences our diet based on needs and desires, providing a possible direction for AI to process information more like humans.

Human behavior is complex, resulting from the interaction between our physiological needs and psychological impulses. While AI has made significant progress in recent years, AI systems do not incorporate the psychological aspects of human intelligence. For example, emotional intelligence is rarely considered part of AI.

"Our main focus is on how to incorporate the emotional aspect of intelligence into AI," said Saptarshi Das, an associate professor of engineering science and mechanics at Pennsylvania State University.

"Emotions are a broad field, and many researchers study psychology. However, for computer engineers, mathematical models and various datasets are crucial for design purposes. Human behavior is easy to observe but difficult to measure, making it challenging to replicate and imbue emotional intelligence in robots. There is currently no real method to do this," Das pointed out.

Das noted that our dietary habits are a good example of the interaction between emotional intelligence and physiological and psychological states. What we eat is largely influenced by the taste process, which refers to how our taste helps us decide what to consume based on flavor preferences. This is different from hunger, which is a physiological reason for eating.

"If you have the luxury of all possible food choices, you would choose your favorite food. You wouldn't choose something very bitter but would instead opt for something sweeter, right?" Das said.

Anyone who has been tempted by a piece of chocolate cake at an afternoon workplace gathering after having a hearty lunch knows that a person can crave something they like even if they are not hungry.

"If you eat something sweet, regardless of your physiological condition, you will eat it, unlike if someone gives you a big piece of meat. Your psychological state still wants to be satisfied, so even if you're not hungry, you will have the urge to eat something sweet," Das said.

Das said that while there are still many questions about the neural circuits and molecular-level mechanisms of hunger perception and appetite control in the brain, advancements in brain imaging provide more information on how these circuits function in terms of taste.

The taste receptors on the human tongue convert chemical data into electrical impulses. These impulses are then sent through neurons to the taste cortex in the brain, a complex network of neurons that shapes our perception of taste. Researchers have developed a simplified biomimetic version of this process, including an electronic "tongue" and an electronic "taste cortex" made of 2D materials.

The artificial taste buds consist of tiny graphene-based electronic sensors called chemFETs, which can detect gases or chemical molecules. Another part of the circuit uses memristors, which are transistors that remember past signals and are made of molybdenum disulfide. This allows researchers to design an "electronic taste cortex" that connects physiological-driven "hunger neurons," psychologically-driven "appetite neurons," and an "eating circuit."

Subir Ghosh, a doctoral student in engineering science and mechanics and co-author of the study, explained that when detecting salt or sodium chloride, the device senses sodium ions.

"This means that the device can 'taste' salt," Ghosh said.

The properties of the two different 2D materials complement each other in creating an artificial taste system.

"We used two different materials because graphene is an excellent chemical sensor, but it is not good for the circuits and logic required to simulate brain circuits," said Andrew Pannone, a graduate assistant in engineering science and mechanics and co-author of the study. "For this reason, we used molybdenum disulfide, which is also a semiconductor. By combining these nanomaterials, we leverage their respective advantages to create a circuit that mimics the taste system."

The technology has broad applications and is suitable for all five major tastes: sweet, salty, sour, bitter, and umami. Das said that such a robotic taste system has great potential applications, from AI-based diets based on emotional intelligence to personalized dining in restaurants. The research team's next goal is to expand the taste range of the electronic tongue.

"We are trying to create a graphene device array that simulates about 10,000 taste receptors on our tongue, each slightly different from the other, allowing us to differentiate subtle differences in taste," Das said. "An example that comes to mind is those who train their tongues and become sommeliers. Perhaps in the future, we can have an AI system that you can train to become a better sommelier."

The next step is to create an integrated taste chip.

"We hope to manufacture the tongue part and taste circuit in a single chip to further simplify it," Ghosh said. "This will be the focus of our upcoming research."

Furthermore, the researchers envision translating the concept of taste emotional intelligence into other senses, such as visual, auditory, tactile, and olfactory emotional intelligence, to aid in the development of advanced AI systems in the future.

"The circuits we demonstrated are very simple, and we hope to increase the capacity of this system to explore other tastes," Pannone said. "But beyond that, we also want to introduce other senses, which will require different patterns and perhaps different materials and devices. These simple circuits can be refined and closely replicate human behavior. Additionally, as we gain a better understanding of how our brains work, it will enable us to improve this technology."