by Daniel Ranard
We're surrounded by algorithms. Facebook uses an algorithm to curate your newsfeed, credit agencies use an algorithm to compute your credit score, and soon an algorithm may replace you in the driver's seat. As algorithms come to dictate larger swaths of life, it's important to understand exactly what they are, and especially how they're drastically changing.
On one hand, algorithms are nothing new. An algorithm is just a precise set of instructions for carrying out a task. Any chef with a cookbook already follows the written algorithms within: add two cups of water, mix until smooth. Unlike computer algorithms, these instructions are written in a language meant for humans, so they still retain some ambiguity: how should you go about lifting the cup, and what does it mean to mix until smooth? While most of us have the know-how to surmount these ambiguities, computers are designed to follow much more precise instructions. And unlike cookbook recipes, computer algorithms don't concern the manipulation of physical objects like cups and bowls, but rather the manipulation of abstract objects like numbers and bits. A computer algorithm might say: "Take two numbers as inputs, multiply the larger one by seven, then add them," and so on. Though a modern computer is designed to convert these instructions into physical manipulations of the electricity within, one might also use an abacus, or pen and paper. Indeed, some of today's algorithms have been around for centuries: the way a computer calculates square roots is not so different from the method prescribed by Hero of Alexandria.
So while we built bigger and better equipment to execute our algorithms in the 20th century, the central idea of the algorithm had remained unchanged for millenia. What would algorithms of the future hold? First, let's take a diversion and explain where we thought the most advanced algorithms were likely headed, decades ago. Then, we'll see how the recent success of machine learning methods has changed that vision for many, even posing new questions about the nature of knowledge and explanation.