Researchers at the Indian Institute of Science have developed tiny molecular devices whose behavior can be tuned in multiple ways, bridging chemistry and computing. These shape-shifting molecules could lead to AI hardware that naturally learns, rather than merely imitating learning.
A new study from the Indian Institute of Science (IISc) suggests that two long-standing efforts in computing — making machines more brain-like and building devices from the molecular level up — may finally be coming together. In a collaboration bringing together chemistry, physics, and electrical engineering, a team led by Professor Sreetosh Goswami developed tiny molecular devices whose behavior can be tuned in multiple ways, potentially opening the door to AI hardware that naturally learns.
Today's leading neuromorphic systems, often based on oxide materials and filamentary switching, function like carefully engineered machines that imitate learning rather than materials that naturally contain it. The IISc team's molecular devices represent a fundamentally different approach — building computing elements from organic molecules that inherently exhibit the adaptive, tunable properties needed for brain-like computation.
“In a collaboration bringing together chemistry, physics, and electrical engineering, a team led by Professor Sreetosh Goswami developed tiny molecular devices whose behavior can be tuned in multiple ways, potentially opening the door to AI hardware that naturally learns.”
The molecular devices can switch between different states in response to various stimuli, and their behavior can be precisely controlled by adjusting the molecular structure. This versatility means a single type of device could potentially perform multiple computational functions, dramatically reducing the complexity and energy consumption of AI systems.
The implications for the AI industry are significant. As artificial intelligence models grow larger and more energy-intensive, there is urgent need for hardware that can process information more efficiently. Molecular computing devices could offer orders-of-magnitude improvements in energy efficiency compared to today's silicon-based chips, while simultaneously providing the adaptive learning capabilities that current hardware can only simulate through software.
Professor Goswami and his team envision a future where AI hardware is not just efficient but truly intelligent at the material level — where the physical substrate of computation learns and adapts just as biological neural networks do. While significant engineering challenges remain before these molecular devices can be manufactured at scale, the research demonstrates a promising new path toward sustainable, brain-inspired computing.
How did this story make you feel?
