John Naughton’s article in The Guardian, “Machine-learning systems are problematic. That’s why tech bosses call them ‘AI’”, provides an entertaining read about a machine learning issue I always focus on when teaching: using the term artificial intelligence or AI only reveals the ignorance of the user, and the inability, as the great Arthur C. Clarke explained in his Third Law, to differentiate a sufficiently advanced technology from magic. As I explain to my students, I use the term machine learning because some time ago I found that when I was talking to some journalists and the term artificial intelligence came up, they would typically illustrate their story with a photo of The Terminator 🙂

Citing other examples of how language affects our perception, particularly in a field like technology where the novelty of the phenomenon often encourages “linguistic engineering”, John mentions terms such as “share”, which in practice means “allow a company to store your preferences” or “accept” when they mean “either say yes or go somewhere else”. But the “artificial intelligence” is even more interesting, because it includes issues we try to pass off as “normal and natural”, when in practice they are simply miscalculations, lack of caution or carelessness.

In short, “artificial intelligence” doesn’t mean “intelligent machine”. Merriam Webster’s definition is:

1. a branch of computer science dealing with the simulation of intelligent behavior in computers
2. the capability of a machine to imitate intelligent human behavior

AI allows us to perform relatively complex tasks subject to a set of rules generally expressed as constraints, allowing us to create programs capable of playing chess, Go, video games, searching a database quickly simulating an understanding of human language to play Jeopardy, or poker, or more recently capable of expressing ideas in the form of text, drawings or video. Does that mean we now have intelligent machines? No. It simply means that by taking a set of data, we can reproduce patterns that generate a particular result. In no way can these machines do more than carry out the specific operations for which they have been programmed. We are a very long way from machines acquiring consciousness. If we ever get there.

The best way to understand machine learning is as advanced statistical processes; sometimes using very complex tools, but that’s all. These types of tools can allow us to make huge progress in the advanced automation of processes, to the point of generating considerable competitive advantages or replacing jobs that were previously performed by people. But that is all we are talking about: advanced statistics that allow us to process situations with an infinite number of variables, to detect anomalies, classify, predict, understand natural language, optimize, etc. Tools with enormous power and applicable to many processes, which in many cases surprise us or even strike fear into us, but simply advanced and applied statistics. Not “intelligence” as such, which has implications of flexibility and plasticity that the human brain typically possesses and algorithms do not, nor “artificial”, in the sense that statistics is a completely natural science and algorithms, for the moment, have not been able to replicate the way in which intelligence is generated in the brain.

Intelligence, in general, has to do with the ability to make decisions in a generalized area. A chess-playing machine can be very good at playing chess, because at its core, chess is a game of probabilistic management of combinatorial spaces, but a “machine” cannot do anything else unless we train it with other data and correctly introduce the corresponding rules or constraints. The human brain’s ability to develop intelligence is due to our neuronal synapses, which scientists have long been trying to replicate to obtain machines that supposedly work like our brain, but faster. But for the moment, any comparison between the so-called “neural networks” and the brain is purely linguistic, and very far from reality and the capabilities to which this could lead.

Artificial intelligence? AI? For the moment, simply an ambiguous, confusing term, which tells us no more than machine learning does. If for no other reason, we should avoid the term AI, because it leads many to believe that a machine can become intelligent as we understand that term as applied to people, with all the potentially negative effects involved. After years of blaming the computer when something doesn’t work, we are now blaming algorithms when people don’t test them properly first. Which is simply a reflection of we humans’ tendency to seek to blame others.

—

This post was previously published on Enrique Dans’ blog.

***

You Might Also Like These From The Good Men Project

You Said ‘Race’, but Are You Actually Talking About Race?

Understanding the Nonbinary: Are You Confusing Gender With Sex?

The Difference Between Compassion for Those With Disabilities & Ableism?

‘Masculinity’ Is Having an Identity Crisis