Indian scientists have developed a thin molecular film that they say can store 16,500 memory states and perform computations, representing a big advance over binary data storage and processing limited to two states that run traditional transistor-based computers.

The researchers at the Indian Institute of Science (IISc), Bangalore, say their molecular film could help slash by hundreds of times the energy and time required to perform certain mathematical operations that play a key role in modern artificial intelligence (AI) systems.

They have designed and tested in their lab the film — a molecular structure crafted out of metals and organic compounds — that can store and process data in 16,500 states, in contrast to the “zero” or “one” state in binary systems and 64 states, the maximum achieved thus far.

“We’ve combined complex physics and chemistry into what we believe could be a groundbreaking technology for the next generation of AI hardware,” said Navakanta Bhat, an expert in silicon systems at the IISc’s Centre for Nanoscience and Engineering.

Bhat and his colleagues at the IISc believe the technology, if successfully advanced into hardware systems, could potentially bring complex AI tasks, such as the training of large language models, to personal devices such as laptops and smartphones.

A paper describing their feat appeared in the journal Nature on Wednesday.

Lack of optimal hardware and associated energy inefficiency are key challenges that currently keep such complex artificial intelligence tasks confined to resource-heavy data centres which are accessible primarily to developers in large corporations, Bhat said.