Agentic AI: Unlike large language models, which respond to specific prompts, agentic systems are designed to perform tasks and make decisions independently.
Algorithm: The set of rules or instructions that a machine (particularly a computer) follows to achieve a goal.
Artificial Intelligence: Imitates the constant processes occurring in human brains and nervous systems, but while our thinking is fed by our senses, AI systems rely on algorithms or machine learning.
ChatGPT: A large language model developed by OpenAI. It is based on the GPT (generative pre-trained transformer) architecture and is designed to understand and generate human-like text based on the input it receives. It was trained using a range of internet content, but the sources have not been publicly disclosed.
Deep learning: An artificial neural network that attempts to mimic human thinking.
Extractive AI system: Recognizes patterns and makes predictions that help it connect related documents and retrieve relevant information from data.
Generative AI (GenAI) system: Uses a large language model (LLM) that consumes vast volumes of data and learns how to identify patterns and structures that it can use to create new images, text, audio and more.
Generative pre-trained transformer: A type of large language model (LLM) that is used for generating human-like text.
Ghost Bibliographic References: Plausible, but FAKE citations generated by ChatGPT, to non-existent research papers (Orduña-Malea, E., Cabezas-Clavijo, Á. ChatGPT and the potential growing of ghost bibliographic references. Scientometrics 128, 5351–5355 (2023). https://doi.org/10.1007/s11192-023-04804-4) see also Hallucinations
Hallucination: A response generated by an AI that contains false or misleading information presented as fact.
Large language model: A type of GenAI specifically architected to understand, summarize and generate text-based content.
Machine learning: Computer programs that can access data and use it to learn for themselves.
Natural language processing (NLP): The ability of computers to understand, interpret and manipulate human language.
OpenAI: An AI research and deployment company and the developer of ChatGPT.
Prompt engineering/design: The process of carefully constructing prompts or instructions for AI language models to help them understand the intent of your input, as well as your desired output.
RAG (Retrieval-Augmented Generation): A technology that enables a large language model to ground its responses in content from an authoritative knowledge base outside of its training data sources.
RAG (Retrieval-Augmented Generation) Fusion: An enhanced RAG approach that combines the results of multiple searches to improve the detail and diversity of an LLM’s response.
Small language model (SLM): A natural language processing (NLP) model that has fewer parameters than an LLM. SLMs are more lightweight, making them suitable for applications where computational resources are limited or where speed is a priority.
Vector search: A search engine that uses a small language model to convert the meaning and context of unstructured data into numbers or vectors and then find other words, topics or categories that are similar. This allows for more nuanced and context-aware outputs.