Adaptivity means the AI's ability to learn and get better at tasks by using new information or adapting to changing situations.
AI, or Artificial Intelligence, is a field of computer science focused on creating computer systems that can perform tasks that typically require human intelligence, such as understanding natural language, recognizing patterns, making decisions, and learning from data.
An algorithm is a step-by-step set of instructions or rules that a computer programme follows to solve a specific problem or complete a task, such as data analysis or decision-making.
Autonomy refers to the ability of an AI tool to perform tasks and make decisions without constant human intervention, relying on its own programmed rules or learned behaviours.
A conversation is a series of interactions between a human user and an AI tool where information is exchanged, questions are answered, or a dialogue takes place. It involves back-and-forth communication, often in a text-based format, with the AI providing responses based on its programming and training.
Data cleaning means identifying and fixing biased or problematic information in the training data to prevent the AI from generating discriminatory, offensive, or harmful content and ensuring the AI tool learns from a reliable and ethical dataset.
Data scraping refers to the process of automatically extracting information or data from websites, documents, or other sources on the internet or from databases. Data scraping tools or scripts are employed to collect, organize, and make use of this data in AI applications.
Deep learning in AI is a technology that uses artificial neural networks with multiple layers to automatically learn and understand patterns and information from data, making it particularly useful for tasks like recognizing images, understanding spoken language, and more.
Echo chambers refer to situations where AI algorithms, such as recommendation systems or content generators, reinforce and amplify existing beliefs or viewpoints by continuously providing content that aligns with them. This can lead to a limited and one-sided information environment, where users are exposed primarily to ideas and opinions they already agree with, potentially isolating them from diverse perspectives.
Fake news refers to false or misleading information created by AI models, often designed to look like real news but is actually fabricated or misleading content.
Filter bubbles are situations where AI algorithms, such as recommendation systems or content generators, show users information or content that aligns with their existing beliefs and preferences, creating a narrow and personalized view of the world. This can limit exposure to diverse perspectives and potentially contribute to echo chambers of information.
Generative AI is a type of artificial intelligence that can create new content, such as text, images, or music, by learning from existing data and mimicking patterns.
GPT stands for ‘Generative Pre-trained Transformer’. It is a type of artificial intelligence model designed for natural language understanding and generation tasks. GPT models are pre-trained on a vast amount of text data and can generate human-like text based on the input they receive.
Hallucination occurs when the AI model generates content that is not based on real information but instead invents details, resulting in inaccurate or fictional output. Hallucination typically happens when the AI lacks proper grounding in factual data or when it extrapolates beyond its training data, leading to misleading or erroneous content.
Jailbreaking means altering the AI model’s configuration or training data to make it generate content that goes against its intended guidelines or ethical standards. This typically happens by tweaking the AI's settings to produce undesirable or harmful output.
Large language models (LLMs) are advanced artificial intelligence systems that have been trained on massive amounts of text data to understand and generate human-like language.
Machine learning is a branch of artificial intelligence where computer systems learn from data to improve their ability to perform tasks without being explicitly programmed.
Narrow AI, also known as Weak AI, refers to artificial intelligence systems that are designed and trained for specific, well-defined tasks or functions, rather than having general intelligence and the ability to perform a wide range of tasks like a human. All current AI is Narrow AI.
Natural Language Processing (NLP) is a branch of AI that helps computers understand and work with human language, like text or speech, for tasks such as translation, chatbots, and more.
In prompt engineering, a persona is a deliberately defined character or identity used to shape the AI model's responses. This persona includes unique traits like tone, style, and background details, and it helps provide context for the AI's generated content. See also Role.
A prompt is a specific instruction or input provided to an AI model to guide it in generating desired output, such as text or content. It serves as a way for users to communicate their intent or request to the AI system.
Prompt engineering involves designing or crafting specific input instructions or prompts to elicit desired responses or content from the AI model. It is the process of formulating prompts to guide the AI system in generating the output you want.
In Chat GPT and similar AI models, regenerate means to create or generate a new response or piece of text. It typically refers to the AI generating a fresh response based on the input or prompt it receives from the user or application. Regenerating allows the AI to generate a different answer when asked the same question multiple times.
A role is a specific function or identity assigned to a component or element within a prompt to direct the AI model in generating content or performing a particular task. Examples of roles might include ‘subject’, ‘action’, or ‘object’, where each role provides a distinct part of the instruction for the AI system. See also Persona.
A task is a specific job or activity that the AI tool is instructed to perform, such as generating text, creating images, or answering questions.
In generative AI, temperature adjusts how random or structured the AI’s output is. Higher values make it more random and creative, while lower values make it more predictable and structured.
A transformer is a specialized technology that not only helps AI understand but also transforms and generates things like text or sequences of data. It plays a crucial role in tasks like language understanding and content creation.
Troll factories are organizations or groups that use AI-powered tools to create and disseminate false or inflammatory content, often with the intention of manipulating public opinion or causing disruption on online platforms. These factories use AI-generated text, images, or other content to amplify their messages and provoke reactions.
Note on Use of Generative AI
This glossary was co-created with ChatGPT (GPT 3.5). The list of terms was human-generated and draws upon knowledge built through engagement with discourse around generative AI as part of the AI Futures project. The definitions were generated by ChatGPT through a series of prompts that requested short and simple definitions of terms we knew were important to the project. You are welcome to read our (opens in a new window)conversation. Small stylistic changes were made to the wording of the definitions. We had sufficient expertise to verify the accuracy of the definitions generated by ChatGPT. Using generative AI for this specific purpose was a low-risk, high-gain exercise because it helped to create a usable resource much more quickly than would have been otherwise possible, and this allowed us to focus our time and energy on tasks that required more critical reflection.