AI glossary

With all new technology comes new words and meanings that can seem overwhelming – or even irrelevant to our work. Knowing the specific artificial intelligence terminology gives us a pathway to better understanding not just what they mean, but also how we can utilise them within our organisations.
We’ve pulled together the following list of key AI terms that you are likely to hear - and what it means for NFPs. This list is designed to help you feel more confident when conversations stray into technical terminology, and some questions you should ask if you hear the terms. This is by no means a complete list of terms but hopefully  it can help with some of the more frequently used terminology.

Update - December 2025:

2025 was a massive year in AI, and with new advancements come new terms! We’ve added new or updated definitions for the following terms:

• AI slop
• Artificial General Intelligence (AGI)
• Artificial Superintelligence (ASI)
• Diffusion models
• Emergent behaviour
• Explainable AI (XAI)
• Generative AI
• Human in the loop (HITL)
• Multimodal AI
• Transformers
   

Artificial Intelligence (AI) 

The simulation of human intelligence processes by machines, especially computer systems. These processes include learning, reasoning, and self-correction. 

AI has been around for a long time but has recently surged into the general conversation through the widespread availability and use of tools like ChatGPT. AI is now embedded in many everyday tools – from writing assistants to spreadsheet analysers – and interacts with users in ways that were previously invisible. This means you’re likely already using AI in tasks like summarising documents, generating reports, or finding insights in data. The questions NFPs should be asking are, should we be considering adding more productivity-generating AI to our toolkit – like a ChatGPT subscription, or even experimenting with custom-built solutions like a chatbot? Questions about costs (what benefits will I get? Will they justify the costs and how do they compare with the benefits/costs of human input?) and ethics (communicating with stakeholders, keeping your data safe, addressing bias) come into play here.  

Agentic AI 

Agentic AI refers to artificial intelligence systems designed to act autonomously, making decisions and taking actions without human intervention. These systems are capable of perceiving their environment, analysing data, and responding to inputs in a way that mimics human agency. Agentic AI can be found in applications such as self-driving cars, autonomous drones, and advanced robotics, where they perform tasks independently and adapt to changing circumstances. Agentic AI has also been explored in programs such as Google’s Gemini and Microsoft365’s Copilot, with the ability to send emails, organise meetings and conduct research on behalf of users.

While still supervised by humans, these systems increasingly make complex decisions independently, raising new considerations for trust and oversight.

Algorithmic Bias 

Systematic and repeatable errors in a computer system that create unfair outcomes, such as privileging one arbitrary group over others. Addressing algorithmic bias is crucial for fair and ethical AI applications. We will talk about some ways to best do this shortly. 

Algorithm 

A set of rules or processes to be followed in calculations or other problem-solving operations, especially by a computer. Like a recipe, but for computers. 

If you don’t know much about algorithms but want to learn more, you’re moving into the worlds of computer and data science, and the Internet is your friend. Otherwise, algorithms are a level of detail that is too much for most people working in NFPs but not in IT.  

AI Ethics 

The stage of AI integration into society in 2025 means that we are preoccupied with considering the moral and ethical implications of AI technologies, including issues such as bias, transparency, accountability, and the impact on employment and society. This is a critical stage and NFPs are uniquely placed to engage and understand the ethical implications of automation that AI involves. 

One of the great contributions that NFPs can make is calling out bias when they see it and helping technicians by contributing ‘domain expertise’ to solving the problem. For example, if a data scientist has analysed a dataset, domain experts can help by asking questions about the data/results, pointing out anomalies, or discussing the ‘real-world’ context for the results.  Check out the AI CARE framework to help your NFP make the right decisions about AI. 

AI Slop

AI slop refers to large volumes of low-quality, generic, or misleading content that is generated by AI and shared widely, often without human review. This can include repetitive articles, poorly summarised reports, inaccurate images, or content created purely to fill space rather than add value.
For NFPs, AI slop presents real risks. It can undermine trust with stakeholders, spread misinformation, and dilute important messages. Organisations should focus on using AI to support high-quality work, with clear review processes and accountability, rather than producing content simply because it is easy to generate.

Read more about identifying AI slop in the wild by joining our Asia-Pacific AI Nonprofit Learning Community.

Artificial General Intelligence (AGI)

Artificial General Intelligence refers to a hypothetical form of AI that can understand, learn, and apply knowledge across a wide range of tasks at a level comparable to a human. Unlike current AI systems, AGI would not be limited to one specific function and could adapt to new situations without being retrained.
AGI does not currently exist, but it is frequently discussed in media, research, and policy debates. These discussions tend to focus on long-term implications, including governance, regulation, ethics, and the societal impacts of highly capable AI systems.

Artificial Super Intelligence (ASI)

Artificial Superintelligence refers to a hypothetical future form of AI that would be more capable than humans across a wide range of tasks, including learning, problem-solving, and decision-making. It is described as a step beyond Artificial General Intelligence, where systems are not just human-like but exceed human abilities.

ASI does not exist today and remains a theoretical concept. It is mainly discussed in academic and policy settings as a way to think about how increasingly capable AI systems might be governed and developed responsibly over time.

Bias 

Bias is an error from using a simplified model to represent a complex real-world issue. 

Bias occurs when AI systems reflect the limitations or imbalances of the data they are trained on. This can result in decisions that favour some groups over others. Modern tools now provide ways to detect and mitigate bias, but human expertise is still needed to interpret results, validate fairness, and ensure decisions align with organisational standards.

Big Data 

Extremely large datasets that can be analysed computationally to reveal patterns, trends, and associations. Big Data is often characterized by the three Vs: volume, velocity, and variety.  

Some examples include social media data, emails, sensor data (eg. cameras monitoring how many people walk into a shopping centre) or bank transactions. This data can’t be analysed by traditional software or systems and needs tools like machine learning to generate insights from it. 

Big data can help solve significant social problems, like helping identify unmet needs in cities, or uncovering gendered media reporting. 

Chatbot 

A software application used to conduct an online chat conversation via text or text-to-speech. Chatbots are used in customer service, information acquisition, and other interactive applications. 

Some NFPs are experimenting with support worker and mental health support chatbots to supplement face-to-face services. 

Data Preprocessing 

Data is the foundation of all AI! Preprocessing data refers to the process of cleaning and transforming raw data into a format that can be used by machine learning algorithms. This may include handling missing values, normalising data, and encoding categorical variables. 

If your NFP is considering building AI tools, then the data you will use is a fundamental issue for you. You should get skilled support from a data scientist or business analyst for this work. This is where the old adage ‘Garbage In, Garbage Out’ is deeply meaningful because if your data is incomplete, messy, or not set up for the task, your results will be garbage! Modern workflows also emphasise documenting data sources and automating quality checks to ensure consistency

If you are working on or considering an AI project, ask questions about the data source, like how big and clean it is and what permissions/consent were obtained if the dataset uses individual-level data.

You can also book a consult with one of our AI experts to discuss your AI project.  

Deep Learning & Neural Networks 

Deep Learning: A subset of machine learning that uses neural networks with many layers (hence "deep") to analyse various factors of data. It's particularly effective for tasks such as image and speech recognition. 

Deep learning is part of chatbots and spam filters so it's part of our everyday AI use. Your NFP should understand it because it can generate problems with bias, privacy, fairness, and transparency. Knowing that deep learning is part of any AI that your organisation is using helps NFPs to manage these critical ethical issues.  Learn more here. 

Neural networks: A series of algorithms that attempt to recognise underlying relationships in a set of data through a process that mimics the way the human brain operates. 

While you may not really need to know much about neural networks, it can be valuable to know if they were used to train the AI you use or develop. Like deep learning, neural networks are often the key methods behind why AI is so useful. But it also suffers the same challenges of learning patterns from data that go on to generate biased responses. For example, if all the datasets are from Western sources, then neural networks produce results that are culturally biased.  

Diffusion Models

Diffusion models are a type of AI used to generate new content such as images, video, audio, and sometimes text. They work by starting with a pattern of random visual or data “static” and gradually refining it into meaningful content based on patterns learned from their training data.
Many modern image and media generation tools use diffusion models to create realistic outputs from text or other inputs. Like other generative AI techniques, diffusion models raise questions about accuracy, representation, and how training data is sourced and used.

Emergent behaviour

Emergent behaviour refers to unexpected actions or capabilities that arise in AI systems as they become more complex, even though those behaviours were not explicitly programmed. This phenomenon is often observed in large models trained on vast and diverse datasets.

Emergent behaviour highlights the difficulty of fully predicting how advanced AI systems will behave once deployed, reinforcing the importance of testing, monitoring, and clear boundaries around use.

Explainable AI (XAI)

Explainable AI refers to methods and systems that make it possible for humans to understand how and why an AI model makes particular decisions. This can include explanations of which data inputs or factors influenced an outcome.

XAI is especially important in contexts where transparency, accountability, and trust are required. It helps users assess whether an AI system is behaving as intended and supports oversight of automated decision-making.

Generative AI

Generative AI refers to AI systems that can create new content, such as text, images, audio, video, or code, based on patterns learned from existing data. Tools like ChatGPT and image generators are common examples.

NFPs often use generative AI to draft documents, summarise reports, create communications, or brainstorm ideas. While powerful, these tools should be used with care, ensuring outputs are accurate, inclusive, and aligned with organisational values.

Hallucinations 

In the context of AI, Hallucinations are answers to prompts that are factually incorrect. . Hallucinations happen when AI generates outputs that are incorrect, fabricated, or misleading. They often result from gaps or biases in training data, unclear instructions, or overly broad model behaviour.

To best prevent hallucinations, you can provide clear guidelines when prompting, use high-quality data that is diverse and free of bias, test your AI model consistently, and make sure that there is a human reviewer able to legitimise any responses provided. 

Human in the loop (HITL)

Human-in-the-Loop refers to AI systems where humans participate actively in the process, reviewing, guiding, or correcting the AI’s outputs. HITL ensures that decisions, predictions, or recommendations generated by AI are monitored for accuracy, fairness, and context. This approach is especially important in high-stakes applications, like healthcare, legal, or automated decision-making, where errors could have serious consequences.

Example: A case management AI might suggest actions based on client data, but a human worker reviews and approves the recommendations before they are implemented.

Internet of Things (IoT) 

A network of physical objects embedded with sensors, software, and other technologies to connect and exchange data with other devices and systems over the internet. 

Climate focused NFPs may use IoT to monitor things like air quality, endangered species or rubbish collection. Cameras in public places to monitor crowds or support public safety are another example of IoT for public good. 

Large Language Model (LLM) 

Large Language Models are AI models that are designed to understand and generate text and other forms of content like a human. They can summarise text, infer information from context and generate contextually relevant responses. This is done using neural networks which are trained on vast amounts of data sets to help the LLM to predict the probability of the next work in a sentence based on the context provided by proceeding words. Programs like ChatGPT, Gemini, CoPilot and other chatbots are considered LLMs. 

Machine Learning (ML) 

A subset of AI that involves the use of algorithms and statistical models to enable a system to improve its performance on a specific task based on data, without being explicitly programmed. 

NFPs can use ML for tasks such as analysing free-text data, like case notes. Many of the platforms used by NFPs already incorporate ML – think of how Microsoft Word suggests alternative words or phrases, based on your writing style, or the personalised results from your Google search. Some NFP examples are found here. 

If you want to check a machine learning model, ask about things like the accuracy of the model. For example, it might be accurate 90% of the time and you are concerned about the 10% when the model is wrong. Check with the ML engineer but also consider this against your existing human accuracy rate. People also make mistakes! 

Multimodal AI

Multimodal AI refers to AI systems that can understand and work with multiple types of data at the same time, such as text, images, audio, and video. For example, a multimodal system might analyse an image and answer questions about it using text.

This capability opens new possibilities for NFPs, such as analysing video or audio data, improving accessibility, or supporting richer service delivery. As with all AI, it also raises considerations around privacy, consent, and appropriate use of sensitive data.

Natural Language Processing (NLP) 

A field of AI focused on the interaction between computers and humans through natural language. It involves designing algorithms to understand and generate human language. 

Some of the AI add-ins that are now available for programs allow users without specific technical skills to do more technical tasks. For example, querying spreadsheets or documents using plain-language questions  (e.g. how many clients are in an age bracket) rather than programming a particular function. This can be amazing, but beware – to quote a superhero’s uncle ‘with great power comes great responsibility’! You need to understand how to frame the right question and how to interpret the results.    

Prompt writing or prompt engineering

Prompt writing refers to the practice of crafting specific inputs that guide an artificial intelligence model to generate desired responses. This can be in the form of questions, statements, or commands that provide clear and concise instructions. This helps the AI understand the context and deliver accurate, relevant, and coherent outputs. The more detail that is provided in the prompt, the more detailed and relevant of an answer you will receive. 

Robotics 

A branch of technology that deals with the design, construction, operation, and application of robots. Robotics often overlaps with AI to create intelligent machines, using large amounts of data and requiring complex analytics to program the robots for their specific tasks. 

Robots are being used in healthcare settings to help staff with mundane tasks and even as support aids for dementia patients or the elderly. 

Transformers

Transformers are a type of neural network architecture that significantly improved how AI handles language and other sequential data. Transformers allow LLM’s to consider all parts of a sentence or document at once, rather than one word at a time. This approach powers many advanced AI systems, enabling tasks such as language translation, summarisation, and question‑answering efficiently.

Our own large language model is ever evolving

New technologies bring new terms, and increasing our understanding of the common words will help us truly utilise AI for our own missions for social good.

The above should help you with understanding the often confusing world of AI. If there are any terms that you think might be worth including in this glossary, comment them below! We'll continue to update this page with more terms, so make sure to check back or bookmark this page for an easy reference point.
You’ve learned the terms, now dive deeper into AI by exploring our upcoming webinars, written articles and self-paced learning through the my learning tab.