Artificial Intelligence

Algorithms*: 

Specific instruction sets or approaches used by models to generate unique content that determines the use of recognized patterns learned in training.

Artificial Intelligence (AI):

The ability of computers to perform tasks that usually require human intelligence, such as learning, problem-solving, and decision-making. 

Bias:

A tendency of an AI system to make unfair or inaccurate predictions or decisions due to prejudiced data or flawed algorithms. 

Data:

Information that computers can process, such as numbers, text, images, and sounds. 

Generative Artificial Intelligence (Generative AI):

A type of AI that creates new content, such as text, images, music, or video, from prompts given to it by humans. It does this using learning patterns from existing data that has been used for its training.  It is important to know that objects created by Generative AI, while original, might not be correct or accurate.  

Hallucination:

When an AI system generates information or outputs that seem real or accurate but are actually fabricated and not based on real data or facts. 

Model:

A mathematical representation of a real-world process that a computer uses to make predictions or decisions based on data.  

Diffusion Models*: Add and refine random bits to match input descriptions from training on labeled images until an image has been created that meets the descriptive input. 

Transformer Models*:  runs parallel queries (rather than sequential) on datasets that it breaks down into smaller portions.  

Generative Pre-Trained Transformer (GPT) models*:  after being trained on large datasets of code or language, these AI systems can generate novel produces (text or code) without minimal prompting by the user. 

Large Language Models (LLMs)*: a subset of transformer models that processes large amounts of language/text to generate original content 

Small Language Models (SLMs)+: similar to LLMs except they operate within a closed or limited system, allowing for a more streamlined, efficient, and faster model that requires less infrastructure/power for computations and training. 

Phantom Citation:

When an AI system generates references or citations to sources that do not exist or cannot be found, leading to false or misleading information. 

Training Data:

The data used to teach a machine learning model, helping it learn to make predictions or decisions. 

Sources: 

OpenAI. (2024) ChatGPT. (July version) [Large language model]. https://chatgpt.com/c/ea038a67-348a-43c9-ac75-7472ab758bc9   

*San Diego State University. (2025). Module 1: How does AI work? Mechanics of text-based and image-based apps. Canvas. https://csuco.instructure.com/courses/2850/modules

+Small language models: The next frontier in enterprise AI. (2024, Jul 26). TECHCiRCLE, Retrieved from https://libproxy.csun.edu/login?url=https://www.proquest.com/magazines/small-language-models-next-frontier-enterprise-ai/docview/3084614117/se-2

Last updated 4/28/25

For handbooks on Artificial Intelligence see OneSearch Artificial Intelligence handbooks. For encyclopedias on Artificial Intelligence see OneSearch Artificial Intelligence encyclopedias.