AI Engineering Solutions from The Garden

AI engineering solutions from the garden

ISE Magazine April 2020 Volume: 52 Number: 4

By  Joseph Byrum


Sometimes inspiration comes from the unlikeliest of places. For Alan Turing, the code-breaking hero of World War II, the patterns he found in nature influenced his approach to solving the problems of computing and artificial intelligence. It all started in his garden.

Last July, the Bank of England announced the renowned computer scientist would appear on the 50-pound bank note opposite the queen, an honor once reserved for 17th and 18th century pioneers like steam engine creator James Watt. The first note will circulate toward the end of 2021.

The idea behind the currency change is to showcase Turing’s work developing the automatic computing engine that helped break Axis ciphers in the 1940s. The Bank of England’s selection committee noted Turing himself said the primitive computing device was “only a foretaste of what is to come, and only the shadow of what is going to be.”

Those achievements were on display in the Oscar winning film The Imitation Game, but don’t expect Benedict Cumberbatch to reprise his role in a sequel (or more accurately, prequel) in which Turing might be found in his garden counting the spirals on the face of sunflowers.

But that’s the sort of thing Turing spent time doing. He looked at the face of the sunflower and wanted to learn why it looked the way it did. While the distribution of seeds on the flower’s face appeared to be random, his counts revealed a complicated pattern. Specifically, the sunflower tended to have 89 spirals of densely packed seeds in one direction and 55 in the other. Both 55 and 89 are numbers in the Fibonacci sequence, which describes a series in which each figure is the sum of the two that came before it. Thus, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55, 89, 144 and so on are Fibonacci numbers. Numbers in this sequence are related to the next number in a ratio that approaches closer to the golden mean, 1.618, as the value of the number increases.

Turing was so fascinated by the discovery that he set out to find a physical explanation for his mathematical observation. He entered the discipline of quantitative biology, which became a lifelong passion. His final published work, “The Chemical Basis of Morphogenesis (1952),” still in-spires researchers today.

In the paper, Turing described how chemicals react and diffuse with one another to create seemingly random-looking structures, like spots on a leopard or stripes on a zebra, that actually follow a discernible mathematical pattern. Biologists and engineers in China recently applied this understanding of the reaction-diffusion process to create a highly effective water filter that promises to have related uses in other fields such as medicine.

The link between biology and AI

There was nothing random about Turing’s interests in biology, cryptology, artificial intelligence (AI) and computing. In each, Turing sought to draw out a discernible, mathematical pattern to create a greater understanding.

The “imitation game,” for instance, was created as a test to determine whether a computer at some point in the future might have the ability to communicate with several humans in such a convincing fashion that the participants wouldn’t know they were talking to a machine. To beat the Turing Test, a machine would “learn” the patterns of humans and replicate them. The machine would gather knowledge in the same way a child does.

He thought it would take 50 years to develop machines powerful enough to do this and win the game, which was not a bad estimate considering how much progress has been made recently.

He saw the computing machines of his time not as crude, unwieldy and extremely limited devices. Turing was looking at the potential he thought would not be realized until the next century. Even if the future machines didn’t “think” the way we do, he reasoned, the result might nonetheless be indistinguishable from actual thought. A machine, despite being constrained by mathematics and logic, might be able to mimic the apparent randomness of human creativity.

As Turing noted, “the appreciation of something as surprising requires as much of a ‘creative mental act’ whether the surprising event originates from a man, a book, a ma-chine or anything else.”

Applying Turing’s method  to the world of finance

Turing never wrote a paper on finance, but he did attempt to apply his genius to maintain his personal wealth. With the uncertainty of war in 1940, he decided the smartest move was to trade his cash for silver ingots. He proceeded to bury them around Bletchley Park as a means of evading the twin threats of foreign invasion and the tax man.

Modern financial analysts have discovered more impressive mathematical techniques to maximize return than burying bullion in the backyard. Traders have found, for instance, that the same Fibonacci sequence that determines a sunflower’s look also plays some kind of role in stock prices. They use a set of numbers in the Fibonacci sequence (23.6, 38.2, 50, 61.8, 78.6 and 100 most commonly) as value levels to consider when trying to forecast the direction of a stock price.

If a stock is on the rise, an analyst will pay attention as it hits each threshold – 23.6 percent, 38.2 percent, and so on. At the upper thresholds like 76.4 and 100, the analyst especially prepares for the stock to reverse course and start losing value. On the downswing, the same intervals become the “floor” for the stock’s value. Once it hits one of those lower intervals, traders believe the stock tends not to go any lower.

Why do these intervals matter? Most believe it is a con-sequence of market psychology. The Fibonacci ratios and golden mean in sunflowers reflect the order and balance of nature. When traders see a commodity rise by, say 100%, they might feel the stock value is disproportionately high – similarly to how things in nature don’t feel right when they are out of proportion.

None of this is necessarily a conscious choice. It’s just a possible explanation underlying a system commonly used in finance and trading algorithms.

Fibonacci ratios show up in the unlikeliest of places, and in so doing, they can serve as an unlikely inspiration. They are, like Turing’s sunflowers, a reminder that order may be found in systems that appear to be inherently disordered – an important lesson for all systems engineers.

Source: IISE Magazine April 2020