If you’re new to the wild world of artificial intelligence, then odds are you’ve come across a dizzying array of new terminology, insider jargon, and words specific to the field.
If this describes you, then you’re in luck. This AI glossary is a one-stop list of the most important terms to know in artificial intelligence, its sub-fields, core technologies, and related disciplines in a format that can be referred back to when you find yourself saying, “wait, what does this mean again?”
Navigating the future of AI means being able to translate, understand, and speak the language of those in the know and this artificial intelligence glossary will help guide you through the vocabulary that matters most.
Already know what you want? Check out the full list of terms here.
AI and Machine Learning Glossary
A set of rules and repeatable steps to be followed (usually by a computer) to solve a problem or carry out a calculation.
Artificial General Intelligence (AGI)
A non-biological intelligence that can accomplish any task at the same level as humans.
Artificial Intelligence (AI)
Artificial Neural Network (ANN)
As the relationship of the name to its biological equivalent suggests, an artificial neural network is basically an algorithm that attempts to replicate the operation of the human brain through the utilization of connected neurons which are organized in layers and send information to each other.
Black Box Algorithms
A black box algorithm is where the result of, or process behind, an algorithm’s decision making is unknown, not understood, or difficult to explain.
A simple way to think of this is teaching machines how to visually interpret the world (teaching them to ‘see’). Perhaps a better way to describe computer vision is as the field of AI and computer science that deals with how machines can gain a level of understanding from images or videos.
Difficult to define but undeniable when it’s present, consciousness is widely regarded as the state of subjective experience – the state or quality of awareness.
Officially ‘Cybernetic Organism’ but usually used to refer to a hybrid of human and machine.
The idea of embodied AI comes from that of embodied cognition which suggests that intelligence is as much a part of the body as it is a part of the brain. With this in mind, embodied AI (for example, bringing sensory input and robotics into the equation) has a beneficial effect on the cognitive function of the AI, allowing it to better understand its situation and surroundings for a more thorough data analysis and response processing.
Explainable AI (X.A.I)
Also known as transparent AI, explainable AI refers to artificial intelligence that carries out actions which are easily understood by humans and can be trusted.
Normally, machine learning tasks like computer vision require the input of massive amounts of image data to train a system, however, the goal of few-shot (and even one-shot) learning is to create a system that greatly reduces the amount of training needed to learn.
Friendly Artificial Intelligence (FIA)
If the values of an artificial general intelligence are aligned with our own, then it is known as friendly AI. In this hypothetical scenario, a friendly artificial intelligence would have a positive benefit on humanity. See also unfriendly artificial intelligence.
The capability to achieve pretty much any goal, including the ability to learn. Should be distinguished from artificial general intelligence which is the ability of AI to accomplish any cognitive task to the same level as humans.
Generative Adversarial Network (GAN)
When two neural networks are trained on an identical set of data, such as images or text, it’s then possible to create what’s known as a generative adversarial network or GAN. This is a class of AI algorithm which is used in unsupervised machine learning.
Another term for artificial intelligence, human-level AI is the point at which a non-biological intelligence is able to complete any cognitive task that humans are capable of.
Intelligence is the ability for an entity (biological or artificial) to accomplish complex goals.
Intelligence explosion describes a scenario where a level of rapidly occurring improvements to general intelligence are on a clear path to AI, reaching superintelligence. It will almost certainly be clear when an intelligence explosion is underway and at this point, the intelligence of AI will rapidly begin to surpass and accelerate beyond even the smartest humans. The recursive nature of an intelligence explosion highlights why it’s so important for us to build safety protocols and safeguards into the development of AI before we reach this point of no-return.
Machine learning is the study of algorithms that improve through experience and is the driving force behind many of the most recent breakthroughs in AI. Learn more about machine learning here.
The ability to accomplish a very specific (narrow) set of goals, tasks, or objectives such as mastering a video game, or driving a car. AI currently falls within the remit of narrow intelligence.
Natural Language Processing (NLP)
In the field of artificial intelligence, NLP focuses on the interactions between computers and human language in both the spoken and written form, in particular through the processing, analysis, and comprehension of large amounts of natural language data. Learn more about NLP here.
Modern terminology uses neural networks to refer to artificial neural networks, however a neural network could also be used to define a biological neural circuit comprised of neurons (the structure of which ANNs are based).
Similar to how a dog is rewarded for successfully carrying out a command, reinforcement learning utilizes a system of rewards and disincentives in the process of machines learning to carry out a new task. As with intelligent biological entities (like us) reinforcement learning starts the machine at a low level and improves performance through learning and continual feedback.
The singularity (also referred to as the technologial singularity) is a term initially coined and popularized by inventor Ray Kurzweil and refers to the point at which we experience an intelligence explosion, where the level of self-improving and learning AI overtakes and rapidly leaves behind our ownn, ultimately reaching superintelligence. It’s hypothesized that the singularity is the point we realize we’ve reached (or are imminently going to reach) a level of artificial superintelligence and can expect a subsequent explosion in technological and scientific growth at a runaway rate. This will almost certainly result in unimaginable changes to human civilization.
Also known as artificial general intelligence, strong AI is any artificial intelligence with the ability to carry out most tasks humans can do, as opposed to weak AI which can only specialize in a specific task at one time.
When an AI reaches a level of general intelligence that massively exceeds our own. This new level of intelligence is known as superintelligence.
This method utilizes labeled data (tagged by a human) to teach a machine learning algorithm how to go about solving a problem or completing a task.
Transfer learning reuses training data from one task to use on another and without the need to retrain the system from the beginning.
Unfriendly Artificial Intelligence
Whether or not a general AI is aligned with our values and can therefore be considered ‘human friendly’ is a key concern when it comes to the ethical implications and existential risks studied in the field of AI development. An unfriendly AI is considered a very possible threat if safeguards aren’t put in place to prevent the development of AI that pursues goals misaligned with our own. See also friendly AI.
A type of machine learning algorithm where machines effectively teach themselves from unlabeled data and without explicit instruction.
Also known as narrow AI, weak artificial intelligence has the ability to only accomplish one thing at a time.
Artificial Intelligence Glossary: A-Z
Artificial General Intelligence
Artificial Neural Network
Black Box Algorithms
Friendly Artificial Intelligence
Generative Adversarial Network