Can a computer learn from the past and predict what will happen next, just like a human? You might not be surprised to learn that some cutting-edge AI models could achieve this feat, but what about a computer that looks a little different – more like a water tank?
We’ve built a small proof-of-concept computer that uses running water instead of a traditional logic circuit processor and predicts future events via an approach called “reservoir computing.”
In benchmark tests, our analog computer performed well at remembering input data and predicting future events—and in some cases even outperformed a high-performance digital computer.
So how does it work?
Throwing stones into the pond
Imagine two children, Alice and Bob, playing at the edge of a pond. Bob throws large and small stones into the water one at a time, seemingly at random.
Large and small stones create water waves of different sizes. Alice watches the water waves created by the stones and learns to predict what the waves will do next – and from that she can get an idea of which stone Bob will throw next.
Reservoir computers replicate the reasoning process taking place in Alice’s brain. They can learn from past input to predict future events.
Although reservoir computers were first proposed using neural networks—computer programs loosely based on the structure of neurons in the brain—they can also be built with simple physical systems.
Reservoir computers are analog computers. An analog computer represents data continuously, unlike digital computers, which represent data as suddenly changing binary “zero” and “one” states.
Representing data in a continuous manner allows analog computers to model certain natural events—those that occur in a kind of unpredictable sequence called a “chaotic time series”—better than a digital computer.
How to make predictions
To understand how we can use a reservoir computer to make predictions, imagine that you have a record of daily rainfall for the past year and a bucket full of water near you. The bucket becomes our “computational reservoir”.
We enter the daily rainfall record into the bucket using stones. For a day of light rain we throw a small stone; for a day of heavy rain, a large stone. For a day without rain, we throw no stones.
Each stone creates waves, which then splash around the bucket and interact with waves created by other stones.
Read more: There’s a way to turn almost any object into a computer—and it could cause shockwaves in artificial intelligence
At the end of this process, the state of the water in the bucket gives us a prediction. If the interactions between waves create large new waves, we can say that our reservoir computer is predicting heavy rain. But if they are small, we should only expect light rain.
It is also possible that the waves will cancel each other out and form a still water surface. In that case, we shouldn’t expect rain.
The reservoir makes a weather forecast because the waves in the bucket and the rainfall patterns develop over time according to the same physical laws.
So do many other natural and socio-economic processes. This means that a reservoir computer can also predict financial markets and even certain types of human activity.
Longer lasting waves
The “gang with water” reservoir computer has its limitations. First, the waves are short-lived. To predict complex processes such as climate change and population growth, we need a reservoir of more durable waves.
One possibility is “solitons”. These are self-reinforcing waves that hold their shape and travel long distances.
For our reservoir computer, we used compact soliton-like waves. You often see such waves in a bathroom sink or drinking fountain.
In our computer, a thin layer of water flows over a slightly inclined metal plate. A small electric pump changes the speed of the flow and creates solitary waves.
We added a fluorescent material to make the water glow under ultraviolet light, to precisely measure the size of the waves.
The pump plays the role of falling stones in the game of Alice and Bob, but the lonely waves correspond to the waves on the surface of the water. Solitary waves travel much faster and live longer than water waves in a bucket, allowing our computer to process data at a higher rate.
So how does it work?
We tested our computer’s ability to remember past input and to make predictions on a benchmark set of chaotic and random data. Our computer not only performed all tasks exceptionally well, but also outperformed a high-performance digital computer with the same problem.
Together with my colleague Andrey Pototsky, we also created a mathematical model that enabled us to better understand the physical properties of the solitary waves.
Next, we plan to miniaturize our computer as a microfluidic processor. Water waves should be able to perform calculations inside a chip that works similarly to the silicon chips used in every smartphone.
Read more: What was the first computer?
In the future, our computer may be able to produce reliable long-term forecasts in areas such as climate change, forest fires and financial markets – at much lower cost and wider availability than current supercomputers.
Our computer is also naturally immune to cyber attacks because it does not use digital data.
Our vision is that a soliton-based microfluidic reservoir computer will bring data science and machine learning to rural and remote communities worldwide. But for now, our research work continues.
#Scientists #built #analog #computer #water #waves #predict #chaotic #future