Artificially intelligent: A brief glossary of the ideas behind AI

Written by Jaymin Shah

What if the scariest thing about AI is the vocabulary test?

Sure, some people worry about machines putting an end to humanity as we know it, but that may be no more likely than a Mayan-predicted apocalypse. Artificial intelligence, though, may actually bring us lots of things we’ll like: self-driving cars, caregivers for the sick, personal assistants that know exactly what you need, when you need it.

Yes, there is math and science, and a lot of it, behind the current boom in AI research, products and services. But you don’t have to have a Ph.D. to get a handle on the basic ideas.

Here’s a quick rundown of some notions you should know.

agent Software that reacts to things happening around it without the direct instruction of a user. A step beyond conventional software programs because they’re always on and work by themselves, agents generally perform a single, specialized task, such as assembling news feeds or ordering email in terms of importance.

algorithm A formula or step-by-step process for a specific task. Think of it as a mathematical recipe or flow chart (“If x = 1, then…”)

artificial intelligence The branch of computer science addressing simulated intelligence in machines. John McCarthy, the man who coined the term six decades ago, defined it as “the science and engineering of making intelligent machines, especially intelligent computer programs.”

deep learning The area of machine learning (see below) designed to recognize relationships in data. It tries to do in software what we do with our brains.

machine learning The aspect of AI that allows computers to learn tasks or to recognize new patterns on their own, without being explicitly programmed to do so. That is, it’s a form of adaptation and an essential ingredient in cooking up true intelligence.

natural language processing The ability of a computer to understand and use human speech as it is spoken. Natural language processing is a key element of AI that’s built off machine learning. If realized, it could even eliminate the need for programming languages.

neural network A form of information processing that uses multiple nodes — anything connected to a network — to emulate the biology of the brain. Neural networks use inputs from the nodes to tackle problems from multiple angles and make inferences from observations, rather than following a set of instructions. The technology is used in tasks such as handwriting recognition, in which common symbols aren’t all rendered alike.

robot Software that simulates a human activity, such as a comparison-shopping program. Often shortened to “bot.” The robots you may be thinking of — clanky, wheezy hardware that walks like a dog, or the bright, shiny C-3PO — involve a lot of mechanical capabilities that are separate from artificial intelligence.

strong AI What we’ll have when artificial intelligence fully rivals, or even exceeds, our own. We’re still a long ways off.

Turing test A challenge proposed by English computer scientist Alan Turing in 1950 for evaluating a computer’s ability to demonstrate intelligent behavior. To pass the Turing test, a computer’s natural language responses would be indistinguishable from a human being’s responses. But the Turing test’s very prescribed and narrow circumstances are a far cry from the more multifaceted, commonsense interaction that AI researchers are looking for.

weak AI What we have now. Limited, single-function software like Google’s AlphaGo program or Facebook image recognition tec

About the author

Jaymin Shah

Jaymin Shah is a tech entrepreneur. He is the Founder & CEO of TechOptimals. He has made a name for himself in the tech media world as a writer relentlessly covering Technology, in addition to a broad range of startups. Contact Jaymin at

Leave a Comment