An algorithm is any sort of routine procedure. A cake recipe is an algorithm for making a cake: if you mix together the ingredients, put them in a pan, put the pan in an oven, and wait for some time, at the end, you will have a cake. Some paper assignments given to students can be done algorithmically: state a thesis, present some arguments, write up a conclusion, and turn it in. The virtue of an algorithm is that it breaks a larger project into a set of smaller steps which, when done in the right order, complete the project. It does not matter who performs the smaller steps so long as they are done correctly. In some cases, the steps are small enough and simple enough for a mindless machine to do them. In that case, when a simple machine can perform an algorithm, we call the algorithm a program. One might think of the laborious “recipe” we follow when we perform long division. That algorithm, or one like it, can be performed by a mindless calculator, thank goodness.
With the advent of computers in the 20th century, more and more of our information has been coded into data that computers can process and manage, which means that more and more of the information at our disposal is processed by algorithms. As I type these words, an algorithm is taking the electrical signals from the keys I press and storing them as numbers; other algorithms are turning those numbers into commands to light up a few pixels on my screen so that I can see what I write. And of course, matters get increasingly complicated from this point forward as I send a file to you through the air and over some wires and your machine receives the signals and you pull up the document and read it. The algorithms in ordinary laptop computers are like complicated factories of routines, all performing their narrow operations so as to produce overall effects that we take for granted—until, that is, something goes wrong and a file won’t load and we curse the machines for being so stupid!
When we turn to the internet to search for the things we are interested in, armies of algorithms take in the information we give them and search for other collections of stored information that “match” what we are looking for (at least according to the programming of the algorithms). The algorithms, of course, do not know what they are doing; they are mindlessly following recipes which (if all goes right) end up with results that satisfy us. Since we cannot count on the algorithms to have any common sense or to know what they are doing, the programmers of the algorithms have to rely on certain tricks that will get the algorithms to do what we want. Search algorithms will look for the websites that most people have ended up going to when they typed words similar to what we typed; they will rank websites according to how popular they are by some measure or other; they may even take a peek at your own history to try to gauge which sites are more likely to satisfy your interests. It is far from foolproof, of course. If you are interested in the historical and cultural background of cockfighting, be careful what search terms you employ or you may be presented with images not strictly relevant to your inquiry.
Exactly how search engines do what they do is a closely guarded secret because search engines are very big business indeed. They are big business because the companies that provide search engines use them as opportunities to provide you with information you did not exactly ask for. This is advertising. If you are interested in baking cakes, you might also be interested in buying special baking pans, a stylish apron, or a new mixer, and advertisements for such products might appear somewhere on your screen. Companies hoping to sell these products pay search engine companies to put those ads on the screens of people who are likely to be interested in the products. The strategy is far more focused and far more effective than placing an ad in a newspaper.
But the business model of search engine companies does not stop there. As you search for items, search engines also gather data about your interests. These data are compiled together with data from all other users so that high-level algorithms can discern larger patterns of human interests and behavior. It may turn out, for example, that people interested in recipes for carrot cake are also more likely to be interested in folksy aprons, and also more likely to be interested in magazines celebrating rural lifestyles. Perhaps they also tend to vote Republican and have pro-life views. Perhaps they also are more likely to buy domestic automobiles and air fryers. I am making up these correlations for the point of illustration, but search engines are actively gathering data to make far more secure assessments as to what sorts of people like what sorts of things. These metadata—or data about the data reflecting people’s online activities—are the real source of wealth for search engine companies. The aim is to know people better than they know themselves. Search engine companies do not exactly sell this information outright, but it plays a central role in a complex, multi-level process in which advertising space is auctioned off to companies in an automated process known as real-time bidding. Basically, real-time bidding is an algorithm that sells access to user’s information to other algorithms so that client companies can mount more effective ad campaigns.
It probably seems to you that this sort of information may be valuable, but it cannot be the most valuable thing in the world. But you are wrong. In 2017, The Economist announced that information had surpassed oil as the world’s most valuable resource. There is more money to be made in gathering information about people’s buying habits than there is in selling them any particular thing.
Algorithms are like vast ant colonies. An individual is fairly simple and robotic in its behavior. But when assembled into great colonies, those simple robots can manage to accomplish extraordinarily complicated tasks. Indeed, they can accomplish any task that can be broken down into simple steps. It practically does not matter how many simple steps need to be performed since we have limitless supplies of “ants” to put to work. The production, storing, searching, filtering, and control of information in the modern world—and even the buying and selling of it—is done primarily by algorithms; the role of humans now is to be sources of information and to consume in an economy based upon it.
There are two broad lessons to be gathered from this discussion of algorithms. The first lesson is that algorithms are mechanical and mindless in the sense that they do not know what they are doing and proceed according to rules in robotic fashion. This means they can make “mistakes” (from the point of view of our own expectations) without realizing it or without anything going wrong in their programming. As we will see, this also means that algorithms can be “gamed” or manipulated in clever ways so as to produce disinformation to users. The second lesson is that there is a lot of power and wealth connected to algorithms and tremendous incentives to find ways to gain control over them. Were Francis Bacon alive today, he would say that “Algorithms themselves are power.”
- Newspaper: a 20th-century artifact made of paper on which was printed news, advertisements, and comics, and delivered to people’s doorsteps; you can find images of such ancient relics on the internet. ↵