Shining a Light on Cellular Automata

Moore's Law, named after Intel co-founder Gordon Moore, is the observation that the number of transistors on a microchip doubles approximately every two years, leading to a significant increase in computational power over time. This principle has been the driving force behind the rapid advancement of technology for several decades. However, the continuation of Moore's Law is increasingly being challenged by physical constraints as components shrink to smaller and smaller sizes.

To prevent slowdowns in the advancement of computer chips caused by the end of Moore's Law, researchers are exploring alternative technologies, such as photonic computing. Photonic computing leverages light, or photons, instead of electrons, to transmit and process data. Light has already been successfully used in the field of communications, allowing for faster data transmission over long distances compared to traditional wired connections. While light can transmit information quickly, it is difficult to manipulate and control light for complex computations. As such, it is still not clear how this technology could replace traditional computer architectures.

It may not be a fully general-purpose computer, but a team of engineers at the California Institute of Technology believe that their photonic computing architecture is a step towards that goal. Their solution is based on the concept of cellular automata, or simulated cells, that follow a predefined set of rules. The most recognizable example of cellular automata is Conway's Game of Life, in which rules that simulate things like overpopulation and reproduction cause the cells to die or flourish. These systems typically run in software on traditional computing hardware, but in this work, a hardware-based cellular automata system was developed that leverages photonics for computation.

One of the factors that makes light-based computing so difficult is that many gates, switches, and other components are needed for transferring and storing light based information around the computer — and such components with sufficient performance do not exist today. This limitation makes cellular automata a desirable architecture. Since this technique only requires that a cell interact with its immediate neighbors, the hardware for controlling light can be greatly simplified. And moreover, the high bandwidth of these photonic connections makes the processing incredibly fast.

The hardware represents elementary cellular automata in a time-multiplexed manner via the pulses of a mode-locked laser with a fixed repetition rate. The presence of pulses represent the state of a live cell, while the absence of pulses represent a dead cell. Cell states are encoded using an electro-optic modulator, which splits the signal between three optical delay lines. These delays allow the signal to interfere with the signals of its nearest neighbors, and thus change the state of the cell that comes before or after it in the one-dimensional lattice. Next, optoelectronic thresholding determines the new state of each cell, and the results are stored on a field-programmable gate array, which serves to drive the next iteration, or cycle, of the computer.

The researchers believe that the ultrafast operations made possible by their system could ultimately lead to the development of a next-generation computer architecture that performs tasks more efficiently than today’s digital computers. However, these cellular automata-based computers are limited in the types of operations that they can perform, so you will probably not have a photonic computer on your desktop any time soon. But even still, the team was able to demonstrate some complex phenomena, including fractals, chaos, and solitons with their device. These sorts of tasks are generally associated with much more complex hardware, so perhaps with some additional work, cellular automata-based systems will become capable of performing a wider range of useful tasks.