Nature-inspired parallel computing

What can a tree teach someone about computers? That's the kind of connection we're about to explore. The opening tells us that Philip Emeagwali solves computing problems by looking beyond familiar approaches and taking ideas from nature. We don't yet know how that works. I'd expect an example next.

First, we need one distinction: sequential computers do calculations one at a time; parallel computers work on calculations simultaneously. Picture a large job divided among many workers instead of one person handling everything in order. More workers could get through it faster, provided they can coordinate their work.

The first example is predicting how oil moves underground. That sounds difficult, and the end of the first paragraph confirms the difficulty: the sequential supercomputers used for this job were too slow and inefficient. Emeagwali's innovation was to tackle it with thousands of connected smaller computers. An oil company could presumably benefit from better predictions.

Why is the calculation so demanding? The second paragraph makes the scale concrete. We need information from tens of thousands of locations, with hundreds of calculations at each location, repeated regularly. Temperature, pressure, viscosity, and the other variables help illustrate the workload. We don't need to memorize that list. We need to understand why doing the calculations one after another was a problem.

Connecting more than 65,000 computers addresses the workload, but it creates another problem: who does what? If the computers interfere with one another, merely having lots of them won't solve the job. Imagine assigning workers without telling them which tasks belong to which person. Adding workers alone wouldn't guarantee useful cooperation.

Here's where the trees enter. Their branching patterns distribute water and sap efficiently and help gather sunlight. Emeagwali uses the mathematical principle behind that structure to design a computer network that gathers and distributes messages efficiently. The oil itself doesn't have to resemble a tree. The useful connection is between an efficient natural structure and the network doing the calculations.

I couldn't build that network from this description, and we aren't expected to. We do need the relationship: a demanding computing problem, a coordination difficulty, and a design inspired by a natural system that helps overcome it. The technical details now have a clear purpose.

The final paragraph offers another example, this time based on honeycomb geometry. Emeagwali claims that this design could predict global weather patterns a century ahead. That's a striking claim; I would want evidence before accepting that it can actually deliver. Keep that reaction separate from what he's saying, though. The author reports his claim, and the shared theme is again a computer design inspired by nature.

Finally, Emeagwali predicts that more computer scientists will look to nature and that this will advance both understanding and technology. That broadens the possible significance of his approach. Our main point should still center on what we've learned about his work: he advances parallel computing by adapting useful principles from natural systems. Oil flow is the developed example; the honeycomb design and future prediction extend that idea.