Machine learning
As experienced explorers of the far reaches of technology, the team at SAPHI have some unique insights into the potential impacts of specific innovations on an economy.
One particular technology frequently utilised in the solutions delivered is machine learning. Ostensibly, machine learning appears to be an impossibly complex subject only accessible to those with a computer engineering degree; however, this article will attempt to break down this complex topic into simpler terms.
What is machine learning?
At a high level, machine learning is nothing more than the science behind making computers learn with minimal human input. The core idea is to develop software that enables these computers to learn from experience and self-improve over time.
From an outsider's perspective, this can seem like magic; or from a more nihilistic perspective, something to fear. However, if we unpack this topic further and peer behind the curtain, what we find is something far more benign and rather easy to manipulate in specific ways to solve some pretty complex problems.
How can this be?
Because, at its heart, machine learning is just a basic algorithm running through a logic loop, consistently returning to a fork of "yes" or "no". Yes to reinforce an action and no to discourage it. Just like how we learn.
To make the algorithm work, we need to fuel it with challenges we term "inputs" that it can learn to recognise and associate with the right actions "outputs". Crudely defined, an input is a thing or scenario that the algorithm encounters, and an output is the decision or action the algorithm initiates in response. The fundamental goal is to train the algorithm to recognise differences between inputs and initiate the associated outputs according to predefined logic.
In the case of SIMPaCT, we want to feed the algorithm a set of scenarios (inputs) and teach it what it should do in response to each one (outputs). For example, teaching it in what scenarios it should increase the amount of water flowing through the irrigation and when it should decrease. If there is a scenario where there are lots of people at the park on a hot day, we want it to recognise that it needs to increase the flow, and when there are only a few there, and the temperature is relatively cool, we want it to decrease the flow to maintain a comfortable ambient temperature.
Overtime, we get the algorithm to a position where it does this specific function automatically and can run without human interaction.
To see more examples of how machine learning works, be sure to check out SAPHI's article "What is Machine Learning? Explained With Dogs"