Ultra energy efficient molecular computer breaks records and blurs boundaries

I’m assuming that your computer uses a good amount of energy, that is if it isn’t a DNA based computer or a super powerful neuromorphic computer – both of which run on almost no energy at all but both of which could soon be the most powerful computing platforms of all time, by a wide wide margin.

Now, to join that club, researchers in the US have made progress on developing another kind of computer, a biocomputer, that has a lot in common with future biological computers, chemical computers, liquid computers, and a technology that the US defence establishment call a Molecular Computer or a Molecular Information System (MIS), that uses molecules to solve problems and which uses 10,000 times less energy than a conventional computer.

If made larger these computers could not only efficiently solve complex problems that normally require a lot of time and energy, but they could also shrink a datacenter the size of a Google hyperscale datacenter into a computing package no bigger than your office desk.

For most of computing history, as chips have decreased in size they have also required less energy to run. But this relationship broke around 15 years ago, meaning that computers that perform large computations aren’t as energy efficient as we might have once hoped.

One way to make future computers use less energy may be to ditch electronics altogether and turn to biology instead. Till Korten at Dresden University of Technology in Germany and his colleagues built a chip based molecular computer that uses molecules travelling through channels to solve problems instead of electrons.

The chip is made of glass and etched in such a way that it encodes a problem for the computer to solve. To perform the computation, the researchers flood the chip with a fluid containing molecules called kinesins and microscopic tubes called microtubules.

Microtubules form part of the inner scaffolding of cells, and kinesins move along them to transport other molecules. The biocomputer design turns this upside down. The microtubules effectively “crowd-surf” on kinesins through the chip’s channels, says Korten. All the microtubules move at once, meaning many calculations can be performed simultaneously.

The microtubules move through the channels and each path one takes corresponds to the computer trying out a solution. The researchers then take an image to read the biocomputer’s output and determine the most successful solution.

Korten says the biocomputer can solve intensive combinatorial problems, similar to the calculations used to figure out the optimal route for an airplane that has to make stops in multiple cities.

His team’s machine solved one such problem that required 128 times more computations than what had previously been considered the state of the art for biocomputers that use the same computational mechanism.

Henry Hess at Columbia University in New York says that this is significant progress compared with the first biocomputers made a decade ago.

A traditional computer would solve this particular problem more quickly, but for combinatorial problems with more variables, electronic computers would eventually need billions of years. While how to speed them up is an open question, for the new biocomputer, researchers would simply add more molecules to make solving the problem in days more feasible, says Dan Nicolau at McGill University in Canada. Here, it is advantageous that microtubules are so small that trillions fit in a gram of fluid.

Surfing molecules also perform each step of computation for 10,000 times less energy than electrons in a traditional computer.

“One way to understand it is that these [molecular] motors have been optimised by a billion years of evolution,” says Korten. They are involved in fundamental processes in living cells so they evolved to be as efficient as possible, he says.

The new device is the most powerful of its kind yet, but it still isn’t advanced enough to be practical. Korten says that for realistic applications, such as in logistics, the team’s computer needs more molecules, which is a manufacturing challenge and leads to more errors caused by microtubules taking wrong turns through the channels.

“We are at the state that electronic computers were in when people just started putting together the first transistors,” he added.

Journal reference: ACS Nanoscience Au, DOI: 10.1021/acsnanoscienceau.2c00013