ITEM: Researchers claim they have built the first hardware for a ‘probabilistic computer’ that use ‘p-bits’ to do what some quantum computers can do, but more cheaply and at room temperature.
Quick but necessary backgrounder: normal computers run on bits that are expressed in 1s or 0s. The reason quantum computers are considered the next big advance in computing power is because they use ‘qubits’, which sport two key properties: (1) superpositioning (i.e. they can potentially represent a 1 or a 0 at the same time) and entanglement (meaning two or more qubits can be permanently connected together to become a system). This is what enables quantum computers to solve far more complex problems than standard computers.
The downside is that quantum computers naturally need specialized hardware, and the qubits need to be supercooled close to absolute zero to keep them stable long enough to produce anything useful (after which they have to be reset).
In 2017, a research group at Purdue University in the US proposed the idea of a probabilistic computer using ‘p-bits’ instead of qubits. Dubbed the “poor man’s qubit”, a p-bit can also be either a 1 or 0 at any given time and fluctuate rapidly between the two – but it can do it with existing computer hardware. That also means p-bits are also easier to stabilize than qubits, so no supercooling is required.
Last week, researchers at Purdue and Tohoku University in Japan revealed a paper (published in Nature) describing how they had successfully constructed p-bits using a modified version of magnetoresistive random-access memory (MRAM), which uses the orientation of magnets to create states of resistance corresponding to 1 or 0. From the Purdue release:
Tohoku University researchers William Borders, Shusuke Fukami and Hideo Ohno altered an MRAM device, making it intentionally unstable to better facilitate the ability of p-bits to fluctuate. Purdue researchers combined this device with a transistor to build a three-terminal unit whose fluctuations could be controlled. Eight such p-bit units were interconnected to build a probabilistic computer.
The research team say they successfully tested the resulting circuit with a “quantum” problem: integer factorization (factoring numbers such as 35,161 and 945 into smaller numbers). While that particular problem could be solved with classic computers, a probabilistic computer can do it using less energy and space, says Ahmed Zeeshan Pervaiz, a Ph.D. student in electrical and computer engineering at Purdue:
“On a chip, this circuit would take up the same area as a transistor, but perform a function that would have taken thousands of transistors to perform. It also operates in a manner that could speed up calculation through the parallel operation of a large number of p-bits.”
Speed is relative
There are, naturally, a couple of drawbacks.
For a start, notes this detailed write-up in Ars Technica, p-bits aren’t that fast (yet). Sampling the output of eight interconnected p-bits to solve one computational problem took 15 seconds – which is hella slow compared to conventional computers. On the other hand, an eight p-bit computer could also potentially handle more complex problems that would crash conventional computers. So if you could connect up a hundred p-bits without a corresponding escalation in the sampling time, a probabilistic computer could still outperform a conventional one.
It might not outperform a quantum computer, but the Purdue/Tohoku team say probabilistic computers can handle subsets of quantum computing problems, from machine learning to optimizing a route for goods to travel to market.
Granted, companies like D-Wave are already targeting optimization apps with a limited form of quantum computing called quantum annealing (watch this video for a good explanation of the difference). But D-Wave’s platform still requires a proprietary platform with supercooled qubits – probabilistic computers could theoretically do what D-Wave does with existing hardware at normal temperatures.
Even so, the other potential drawback is a commercial one. Simply put, probabilistic computers don’t have the commercial backing that quantum computing has. Almost everyone who’s anyone in Big Tech, as well as dozens of start-ups, are developing quantum computer tech. Several of them have functioning quantum computers that are already being used by researchers and developers, albeit mainly to figure out how to build better or more stable quantum computers, or at least how to monetize them.
By comparison, the Purdue/Tohoku is backed mainly by DARPA, Semiconductor Research Corp, Japan’s Council for Science, Technology and Innovation, and the Japan Society for the Promotion of Science.
On the other hand, quantum computers are a long way from the commercial mainstream in large part because of the specialized hardware. Indeed, there are currently over half a dozen proprietary hardware approaches being put through their paces. If tech companies decide p-bits are a viable and cost-effective stopgap solution, the commercial ramp-up to probabilistic computers might be much faster.