Defining Artificial Intelligence: A Glossary of Key AI Terms

Source: istockphoto

A

Algorithm: An algorithm is a sequence of explicit, step-by-step instructions that enables a computer to problem solve.

Artificial General Intelligence (AGI): Also called superintelligence, it’s when the capabilities of machine intelligence is equal to or greater than human intelligence.

Artificial Intelligence (AI): The field of computer science in which machines are enabled to simulate human cognition and learning.

Artificial Narrow Intelligence (ANI): Refers to AI that is limited to a specific set of topics and capabilities.

Artificial Neural Network (ANN): A model used in AI, it is loosely based on the human brain. It consists of neural layers that are used for machine learning.

B

Backpropagation: Also known as “backward propagation of errors,” it is a supervised learning technique where errors are computed at the output and distributed backward through the layers of the artificial neural network. It’s a common method of training an artificial neural network where the system’s initial output is compared to the desired output, then the system is adjusted until the difference is minimized.

C

Convolutional Neural Network (CNN): It’s a type of neural networks used to identify and analyze images.

D

Deep Learning: A machine learning method, consisting of a many-layered artificial neural network. Uses many layers of nonlinear processing to extract features from the data, and then transform the data into different levels of abstraction. It can be supervised, semi-supervised, or unsupervised. Used in speech recognition, computer vision, natural language processing, and pattern recognition.

E

Expert System: An AI system that uses a knowledge base of human domain expertise for problem solving.

F

Forward Chaining: A method where AI looks back and analyzes the rule-based system to find the “if” rules, and to determine which rules to use to find a solution.

G

Generative Adversarial Networks (GAN): A type of AI algorithm used in unsupervised machine learning where there are two neural networks (generator and discriminator) trained on the same data set. The generator produces output, and the discriminator compares the output produced with the original data set in efforts to determine which images are authentic. Based on those results, the generator adjusts its parameters for creating new output. This process is iterated until the discriminator is no longer able to distinguish the generator’s output with the original data set. Used to create photorealistic images.

H

Heuristics: Common-sense rules based on experience. In heuristic programming, programs are self-learning, and improve with experience. Frequently used with expert systems.

I

Inductive reasoning: A logical process where multiple premises that are true or true most of the time, are combined to form a conclusion. Often used in prediction and forecasting.

M

Machine Learning: A subset of AI. Computers algorithms learn from identified patterns in data, and adjust their actions accordingly, without explicit programming.

N

Natural Language Processing (NLP): Applying computer algorithms to determine properties of natural human language in efforts to enable machines to comprehend spoken or written language.

Neural Network: See “Artificial Neural Network.”

R

Reinforcement Learning: A type of machine learning method inspired by behavior psychology. The reinforcement learning algorithm (agent) learns by interacting with its environment and is either penalized or rewarded. The agent seeks to make decisions to maximize reward over time.

S

Strong AI: See “Artificial General Intelligence.”

T

Turing Test: A test of a machine’s ability to behave in an intelligent manner that is indistinguishable from human behavior. Developed by Alan Turing in 1950.

Copyright © 2018 Cami Rosso. All rights reserved.