The Next Generation of Computing: P Computers

The rise of artificial intelligence (AI) and machine learning (ML) has created an IT crisis and a significant need for more hardware that is both energy efficient and scalable. A key step in AI and ML is making decisions based on incomplete data, the best approach for which is to generate a probability for each possible answer. Current classical computers are not able to do this in an energy-efficient way, a limitation that has led to the search for new computing approaches. Quantum computers, which run on qubits, can help meet these challenges, but they are extremely sensitive to their environment, must be kept at extremely low temperatures, and are still in early stages of development.

Kerem Camsari, assistant professor of electrical and computer engineering (ECE) at UC Santa Barbara, believes that probabilistic computers (p-computers) are the solution. P computers are powered by probabilistic bits (p-bits), which interact with other p-bits in the same system. Unlike conventional computer bits, which are in the 0 or 1 state, or qubits, which can be in multiple states at once, p-bits fluctuate between positions and operate at room temperature. Camsari and his collaborators discuss their project which demonstrated the promise of p-computers.

“We have shown that inherently probabilistic computers, built from p-bits, can outperform state-of-the-art software that has been in development for decades,” Camsari said.

Camsari’s group collaborated with scientists from the University of Messina in Italy, with Luke Theogarajan, vice chairman of UCSB’s ECE department, and with physics professor John Martinis, who led the team that built the world’s first quantum computer to achieve quantum supremacy. Together, the researchers have obtained promising results using classical hardware to create domain-specific architectures. They developed a unique parsimonious Ising machine (sIm), a novel computing device used to solve optimization problems and minimize power consumption.

Camsari describes the sIm as a collection of probabilistic bits that can be thought of as people. And each person only has a small group of trusted friends, which are the “scattered” connections of the machine.

“People can make decisions quickly because they each have a small group of trusted friends and they don’t have to hear everyone in a whole network,” he explained. “The process by which these agents reach consensus is similar to that used to solve a difficult optimization problem that satisfies many different constraints. Sparse Ising machines allow us to formulate and solve a wide variety of these optimization problems using the same hardware.

The team’s prototype architecture included a field-programmable gate array (FPGA), powerful hardware that offers far more flexibility than application-specific integrated circuits.

“Imagine a computer chip that lets you program the connections between p-bits in a network without having to make a new chip,” Camsari said.

The researchers showed that their sparse architecture in FPGAs was up to six orders of magnitude faster and had increased sampling rates five to eighteen times faster than those achieved by optimized algorithms used on conventional computers. .

Additionally, they reported that their sim achieves massive parallelism where flips per second – the key figure that measures how quickly a p-computer can make an intelligent decision – scales linearly with the number of p-bits. Camsari refers to the analogy of trusted friends trying to make a decision.

“The key issue is that the process of reaching consensus requires strong communication between people who continually talk to each other based on their latest thoughts,” he noted. “If everyone makes decisions without listening, consensus cannot be reached and the optimization problem is not solved.”

In other words, the faster the p-bits communicate, the faster a consensus can be reached, which is why increasing flips per second, while making sure everyone is listening to each other, is crucial.

“That’s exactly what we achieved in our design,” he explained. “By making sure everyone listened to each other and limiting the number of ‘people’ who could be friends with each other, we parallelized the decision-making process.”

Their work has also shown an ability to scale p-computers up to five thousand p-bits, which Camsari sees as extremely promising, while noting that their ideas are only one piece of the p-computer puzzle.

“For us, these results were just the tip of the iceberg,” he said. “We used existing transistor technology to emulate our probabilistic architectures, but if nanodevices with much higher levels of integration were used to build p-computers, the benefits would be enormous. This is what makes me lose sleep.

An 8 p-bit computer that Camsari and his collaborators built while he was a graduate student and postdoctoral researcher at Purdue University first showed the device’s potential. Their article, published in 2019 in Naturedescribes a ten-fold reduction in energy and a hundred-fold reduction in surface footprint required compared to a typical computer.

“The initial findings, combined with our latest results, mean that building p-computers with millions of p-bits to solve optimization or probabilistic decision-making problems with competitive performance may simply be possible.” , Camsari said.

The research team hopes that one day p-computers will deal with a specific set of naturally probabilistic problems much faster and more efficiently.