TL;DR: Research in both biocomputing and neuromorphic computing may hold the key to better computer energy efficiency. By drawing inspiration from nature’s own efficient systems, such as the human brain, we may be able to address the growing energy demands of our increasingly digital world.

As computers consume more and more electricity, scientists are turning to an unlikely inspiration for greater sustainability: the humble biological cell. This approach, known as biological computing, could slash energy consumption in computational processes.

A recent article in The Conversation highlighted this concept, which draws on nature’s own efficient systems to tackle one of the most pressing challenges in modern computing. As data centers and household devices gobble up roughly 3% of global electricity demand, with artificial intelligence poised to push that figure even higher, the need for energy-efficient alternatives has never been more urgent.

The concept of biological computing is rooted in a principle introduced by IBM scientist Rolf Landauer in 1961. The Landauer limit states that a single computational task, such as setting a bit to zero or one, requires a minimum energy expenditure of about 10⁻²¹ joules (J). While this amount seems negligible, it becomes substantial when considering the billions of operations computers perform.

Operating computers at the Landauer limit would theoretically make electricity consumption for computation and heat management inconsequential. However, there’s a significant catch: to achieve this level of efficiency, operations would need to be performed infinitely slowly. In practice, faster computations inevitably lead to increased energy use.

Current processors operate at clock speeds of billions of cycles per second, using about 10⁻¹¹J per bit – approximately ten billion times more than the Landauer limit. This high-speed operation is a result of computers working serially, executing one operation at a time.

To address this energy dilemma, researchers are exploring a fundamentally different computer design based on massively parallel processing. Instead of relying on a single high-speed “hare” processor, this approach proposes using billions of slower “tortoise” processors, each taking a full second to complete its task. This could theoretically allow computers to operate near the Landauer limit, using orders of magnitude less energy than current systems.

One promising implementation of this idea is network-based biocomputation, which harnesses the power of biological motor proteins – nature’s own nanoscale machines. This system involves encoding computational tasks into nanofabricated mazes of channels, typically made of polymer patterns deposited on silicon wafers. Biofilaments, powered by motor proteins, explore all possible paths through the maze simultaneously.

Each biofilament is just a few nanometres in diameter and about a micrometer long, acting as an individual “computer” by encoding information through its spatial position in the maze. This architecture is particularly suitable for solving combinatorial problems, which are computationally demanding for serial computers.

Experiments have shown that such biocomputers require between 1,000 and 10,000 times less energy per computation than electronic processors. This efficiency stems from the evolved nature of biological motor proteins, which use only the energy necessary to perform their tasks at the required rate – typically a few hundred steps per second, a million times slower than transistors.

Significant progress has been made in this field recently. Heiner Linke, Professor of Nanophysics at Lund University and author of the article in The Conversation, also co-authored a 2023 paper that demonstrated the possibility of operating a computer near the Landauer limit. This breakthrough brings us closer to realizing the potential of ultra-low-energy computing.

While the concept of biocomputation is promising, challenges remain in scaling up these systems to compete with electronic computers in terms of speed and computational power. Researchers must overcome obstacles such as precisely controlling biofilaments, reducing error rates, and integrating these systems with current technology.

If these hurdles can be surmounted, the resulting processors could solve certain types of challenging computational problems with a drastically reduced energy cost. This breakthrough could have far-reaching implications for the future of computing and its environmental impact.

As an alternative approach, researchers are also exploring neuromorphic computing, which attempts to emulate the highly interconnected architecture of the human brain. While the basic physical elements of the brain may not be inherently more energy-efficient than transistors, its unique structure and operation offer intriguing possibilities for energy-efficient computing.