AI Glossary and References

Reference list and diagrams of common terms related to AI (artificial intelligence) and ML (machine learning), with two key takeaways about why AI and ML matter to us all.

Aug 10, 2024

AI usually stands for “Artificial Intelligence”. The term often gets used in inconsistent ways, though. The goal of this Glossary is to serve as a quick reference to help folks make sense of what they read and hear about “AI”.

Audience for this post: Anyone. Zero prior knowledge of AI is required.

Types of Artificial Intelligence

Within the world of AI, we hear a lot of acronyms and terms. Here are some of the most common terms you’ll hear.

AGI — Artificial General Intelligence (future AI that’s comparable to human intelligence, sometimes called “strong AI”)
AI — usually “Artificial Intelligence” (but it’s an overloaded term, and sometimes it gets used to describe products that don’t really have “AI inside”)
ASI — Artificial Super Intelligence (future AI that exceeds human intelligence in all aspects, including problem-solving, creativity, and social skills)
[A]NN — [Artificial] Neural Networks
DL — Deep Learning (e.g. a Deep Neural Network with > 3 layers)
GenAI — Generative AI (creates new content)
GEO — Generative Engine Optimization (the GenAI counterpart of SEO, Search Engine Optimization)
ML — Machine Learning
NLP — Natural Language Processing (sometimes, but not always, involving AI)
RL[HF] — Reinforcement Learning [with Human Feedback]

Figure 1 shows how these terms relate to each other. We’ll briefly cover them all in AI Fundamentals article #1. (For a quick comparison of AI, AGI, and ASI, see this post by NeuralBuddies.)

🗝️ KEY TAKEAWAY: Generative AI is only a small subset of everything that can be considered “AI”. When you read announcements or criticisms of “AI”, be sure to ask whether they are talking about AGI, just the GenAI subset, or other types of AI. It makes a difference in terms of capabilities, risks, and ethics.

**Figure 1 - Artificial Intelligence. Original diagram by Karen Smiley, 2024. Licensed under CC BY-NC-ND 4.0.**

Types of AI Models

Texts in white in Figure 1 are specific kinds of AI models, including ML and GenAI. If you’re not a data scientist or a software developer working on building an AI-based tool, don’t worry too much about understanding the details and differences. AI Fundamentals #2 will cover what you need to know as a potential user of AI tools based on these models. Here’s what the terms mean:

CNN — Convolutional Neural Network (special type of NN model; commonly used for OCR (Optical Character Recognition) and other purposes)
Diffusion models (used for images and music - examples: DALL-E, Riffusion)
GAN — Generative Adversarial Network (set of special NNs used for genAI, such as for images and video)
GPT — Generative Pre-trained Transformer (examples: ChatGPT, GPT-4, Copilot, BERT, Bard, Midjourney)
LLM — Large Language Model (a form of generative AI, often using a GPT, that is trained on a huge body of texts and has billions of parameters)
RAG — Retrieval-Augmented Generation (an information retrieval method sometimes used together with a LLM to focus on a specific set of source data)
RNN — Recurrent Neural Network (type of NN ML that is often used for Text To Speech, or TTS)
VAE — Variational AutoEncoder (often used for image recognition, natural language processing, and anomaly detection)

Figure 2 shows how all of the different “domains” where generative AI can be used. Text, images, speech, music, and video are all possible. So are other applications you might not think of, including writing software source code or generating synthetic data for training an AI-based tool.