• A leading Charles Sturt University artificial intelligence (AI) researcher has demonstrated a new form of ‘accessible AI’
• He has built the prototype of an analog computer that uses running water to predict weather, bushfires, climate variation, global financial markets and other complex and chaotic events and processes that are difficult to predict
• The new computer uses ‘soliton’ waves to remember past inputs and to produce the future forecast, and the prototype can be miniaturised to make a liquid-based microchip that can work everywhere

A leading Charles Sturt University artificial intelligence researcher has developed an analog computer that uses running water and can predict weather and financial markets and most things that have complex and chaotic dynamic behaviour.

Dr Ivan Maksymov (pictured, inset) is a Principal Research Fellow in the Charles Sturt Artificial Intelligence and Cyber Futures Institute, and an Australian Research Council Future Fellow.

“From the point of view of physics, I am the first to experimentally demonstrate that ‘solitons’, a special type of waves, can be used for this kind of computations,” Dr Maksymov said.

“There have been several theoretical predictions of soliton-based computers published in the top journals, but I am the first to build it using a special type of solitons.”

Dr Maksymov makes the point that, from his experience, if a company wants to create an AI system it needs to pay to large corporations like Google, AWS, IBM etc, and use a cloud-based service accessible via the Internet.

“What I have built is a prototype that can be miniaturised to make a liquid-based microchip that can work everywhere,” he said.

“The public may be interested because computational power that is higher than that of a supercomputer can be achieved literally using a kitchen tap, thus creating ‘accessible AI’.”

Dr Maksymov’s paper detailing his work has been recently accepted for publication in Europhysics Letters (also called  EPL), a top physics journal in Europe.

The development of this computer is a part of Dr Maksymov’s endeavour to develop novel unconventional computer architectures and accessible AI systems that will serve the needs of rural and remote communities.

Dr Maksymov’s colleague, applied mathematician Dr Andrey Pototsky at Swinburne University of Technology, also participated in this research project.

According to Dr Maksymov the problems of modelling climate change, forecasting bushfires and predicting global financial markets are difficult to resolve since physical and socio-economic processes that underpin these events and phenomena are complex and often chaotic.

“Therefore, specialised software that requires supercomputers needs to be used to produce credible forecasts,” he said.

“However, supercomputers are very expensive, difficult to maintain and also consume significant electrical power. “Obviously, with only two truly powerful high-performance supercomputers available in Australia - one in Canberra, and another one in Perth - the access to them in regional Australia is limited.

“Yet, the immense computational power of a supercomputer often falls short when chaotic natural phenomena such as storms and tsunamis need to be predicted with high accuracy.”

He said in the last century, the lack of computational power was circumvented using analog computers such as the famous Ishiguro Storm Surge Machine and The Water Integrator that was used to design the largest water irrigation canals in the world.

“An analog computer uses a continuous flow of electric current to present data compared with digital computers that use abrupt binary changes in the current flow that correspond to ‘0’ and ‘1’ states,” he said.

“At the end of the last century analog computers were decommissioned due to the development of powerful digital computers.

“However, just several decades later scientists and engineers realised that analog computers remain very useful for applications in AI, where the complexity of machine learning models grows at an exponential rate, doubling in size every four to five months, thus making traditional digital computers struggle to deliver the necessary performance at low power with adequate cost.”

Dr Maksymov explains how his prototype computer works, noting that running water is ubiquitous in everyday life, but very often we do not pay attention to interesting spatial patterns that it forms.

“An example is a drinking fountain, where water that flows over the inclined metal surface may form a special kind of waves called solitons,” he said.

“Those solitons are similar to the well-known ocean swell waves. (For example, swell waves appear at the beginning of the 2022 COLORBOND TV commercial.) The new computer uses soliton waves to remember the past input and to produce the forecast future.”

The solitons are created by a small electric pump connected to a small reservoir with tap water and they move on the surface of an inclined metal plane. For example, the input electric signal driving the motor can be a set of data that correspond to temperature, humidity and wind speed measured by a meteorological station over the last two months. If we plot such data on a graph, we will see that the resulting curve has a rather chaotic pattern.

“Therefore, solitons created using these data encoded as an electric signal that drive the pump will also have chaotic behaviour. If we let the solitons flow further down the inclined plane, they will create a new chaotic pattern,” Dr Maksymov said.

“Measuring that pattern with a slow-motion digital camera and interpreting the resulting data as temperature, humidity and wind speed, we can obtain a weather forecast for the next two or three days.

“Currency exchange rates, electricity consumption, appearance of dark spots on the Sun and variation in the relative quantity of mature cells in the blood can be predicted in a similar manner.”