AI mimics fruit flies to learn without forgetting
A new AI algorithm mimics fruit fly brains to learn without forgetting old data, solving a major AI problem. This could improve robots and medical diagnostics by enabling continuous learning.
Scientists have developed a new algorithm inspired by the fruit fly’s brain that learns quickly and retains old information without losing past knowledge, a long-standing challenge in AI known as catastrophic forgetting.
The breakthrough, detailed in recent research, mimics how fruit flies use a process called sparse coding to efficiently encode smells in their tiny brains. Unlike most AI systems that overwrite old data when learning new tasks, this method keeps prior memories intact—critical for applications like autonomous robots or medical diagnostics where continuous learning is vital.
The algorithm works by activating only a small subset of neurons for each scent, much like a fly’s olfactory system. This efficiency prevents new information from disrupting existing memories. Early tests show it outperforms traditional neural networks in retaining knowledge over time, a step toward AI that learns incrementally like humans.
Researchers say the next step is scaling the approach for real-world use, potentially revolutionizing fields where AI must adapt without forgetting—such as self-driving cars or personalized healthcare. The work highlights how biology can solve some of AI’s toughest problems.
Read Full Story at Ars Technica →
