Summary: Scientists have discovered how animals differentiate between similar scents. Certain neurons consistently identify different smells, while others respond unpredictably, helping to distinguish nuanced aromas over time. This research, inspired by fruit fly studies, has potential implications for enhancing machine-learning models.
By introducing variability, artificial intelligence (AI) systems may mirror the discernment found in nature.
Key Facts:
- The study identified two types of neurons: “reliable cells” that identify distinct odors, and “unreliable cells” that help differentiate similar scents through experience.
- The variability in neural response was found to originate from a deeper brain circuit, indicating its significant purpose.
- This neural variability could benefit continual learning systems in AI, making them more discerning.
Source: CSHL
Order wine at a fancy restaurant, and the sommelier might describe its aroma as having notes of citrus, tropical fruit, or flowers. Yet, when you take a whiff, it might just smell like … wine. How can wine connoisseurs pick out such similar scents?
Cold Spring Harbor Laboratory (CSHL) Associate Professor Saket Navlakha and Salk Institute researcher Shyam Srinivasan may have the answer. They have found that certain neurons allow fruit flies and mice to distinguish between distinct smells.
The team also observed that with experience, another group of neurons helps the animals differentiate between very similar odors.
The study was inspired by research conducted by former CSHL Assistant Professor Glenn Turner. Years ago, Turner noticed that when exposed to the same scent, some fruit fly neurons consistently fired, while others varied from trial to trial.
At the time, many researchers disregarded these differences as background noise. However, Navlakha and Srinivasan wondered if these variations served a purpose.
“There were two things we were interested in,” Navlakha explains. “Where is this variability coming from? And is it good for anything?”
To address these questions, the team created a fruit fly smell model. The model revealed that the variability originated from a deeper brain circuit than previously believed, indicating its meaningful nature.
Furthermore, the team observed that some neurons responded differently to two highly dissimilar odors, but similarly to similar smells. These neurons, known as reliable cells, helped flies quickly distinguish between different odors.
Another larger group of neurons responded unpredictably when exposed to similar smells. These neurons, referred to as unreliable cells, contribute to the ability to identify specific scents, such as those found in a glass of wine, through learning.
“The model we developed shows that these unreliable cells have utility,” Srinivasan explains. “However, it requires repeated learning experiences to take advantage of them.”
This research extends beyond wine enthusiasts. Srinivasan suggests that the findings could help explain how we learn to differentiate between similar sensations detected by other senses, as well as how we make decisions based on sensory inputs.
Furthermore, the results could lead to improved machine-learning models. Unlike fruit fly and mouse neurons, computers typically respond consistently to the same inputs.
“Perhaps machine-learning models shouldn’t represent the same input in the same way every time,” Navlakha elaborates. “In more continual learning systems, variability could prove beneficial.”
Therefore, this research has the potential to contribute to the development of more discerning and reliable AI systems.
About this olfaction and neuroscience research news
Author: Samuel Diamond
Source: CSHL
Contact: Samuel Diamond – CSHL
Image: The image is credited to Neuroscience News
Original Research: The findings will appear in PLOS Biology