Glossary

3-D printing
Part of a process known as additive manufacturing, where an object is created layer by layer.4th Industrial Revolution
Ongoing transformation in the economy and in how we work and live, driven by a range of new technologies, including cloud computing, big data, data analytics, process automation, artificial intelligence, blockchain, internet of things, mobile technologies, and biotechnology.4 Vs
A framework for characteristics of big data. The four Vs are volume, velocity, variety, and veracity.

5-As framework
The basic activities of the ‘information to impact’ framework created by the AICPA and CIMA. Steps consist of assembling information, analysing for insights, advising to influence, applying for impact, and using acumen. See diagram below.
5th Industrial Revolution
The emerging phase of industrialization, in which we work alongside advanced technology and AI to engage in new daily routines and workplace practices, based on human centricity, resilience, and sustainability.

5-As Diagram

accountability framework
Rules that determine what is the best course of action and who is responsible for taking measures when an AI system causes harm or makes an incorrect decision. It is crucial for the ethical deployment of AI.

accessibility
Ensuring that AI is implemented in a way that is beneficial to all segments of society, including underserved and marginalised communities. This requires investing in education and infrastructure to bridge the digital divide. AI should not exacerbate existing inequalities.

additive manufacturing
A manufacturing process in which an object is created layer by layer. It allows complex parts and components to be produced much cheaper and faster and in an entirely customisable manner. See also 3-D printing.AIC triad
An approach to cybersecurity. The model is aimed at helping organisations understand information security and set up policies to help protect the organisation. AIC is an initialism for what are seen as the most important elements of cybersecurity:

Availability: Put simply, systems must be online and available; otherwise, organisations cannot do business.

Integrity: Making sure that people who modify data are authorised to do so means the data is more likely to be accurate and trustworthy.

Confidentiality: When data is being stored and when it is in use or in transit, there need to be rules in place to limit access to those who are authorised to use the data.
Also known as the CIA triad.

algorithm
A set of rules (instructions or steps) that are designed to carry out an operation, perform tasks, solve a problem, or answer a question.

antivirus software
Software that aims to block and detect viruses in a computerised system or application. Some antivirus software can block activity and access when a virus is detected, as well as inform affected parties, applications, and organisations.
See also virus.application attack
Malicious malware that targets a particular application (app). An example is a denial-of-service (DoS) attack. See also malware and denial-of-service (DoS) attack.artificial intelligence (AI)
A ‘system’s ability to correctly interpret external data, to learn from such data, and to use those learnings to achieve specific goals and tasks through flexible adaptation.'¹
AI is also defined as technology that enables computers and machines to simulate human intelligence (learning, reasoning, problem-solving, and decision-making).²artificial general intelligence (AGI)
AI that possesses the ability to understand, learn, and apply knowledge across a wide range of tasks at a level that is comparable to human intelligence.It remains a theoretical concept with debates about its practical achievability.AI accuracy
AI accuracy evaluates the predictive capability of an AI model. Accuracy is the percentage of predictions the model got right. AI models are trained to learn patterns so they can respond to queries, including those linked to unseen data (data that the model has not encountered during its training).³AI agent
AI agents collect and process data to perform tasks on their own and achieve the goals set by humans. They can make decisions, solve problems, and adapt to new information without needing constant human help. Examples of AI agents are self-driving cars, customer service chatbots, and virtual assistants like Siri and Alexa. AI agents possess several key characteristics:

Social ability: The capacity to communicate and collaborate within their environment.

Autonomy: The ability to operate independently and make decisions without external help.

Reactiveness: The ability to respond immediately to changes in their environment.

Proactiveness: The capacity to take initiative and make decisions to achieve their objectives.

AI exposure
The extent to which individuals, organisations, or systems interact with or are influenced by artificial intelligence technologies. AI exposure also refers to the potential for AI to automate tasks, considering both the technical feasibility and economic attractiveness of such automation. AI system life cycle
The AI system life cycle includes design and development (requirement analysis, data collection, training, testing, integration), installation, deployment, operation, maintenance, and disposal. Agile and rapid prototyping methodologies are used to manage the complexity of the design and development stage. ⁴artificial neural network (ANN)
A computational model inspired by the way biological neural networks in the human brain work. ⁵assistive technology
Hardware or software added to a system to increase accessibility. Designed to assist people with disabilities to carry out daily activities.⁶augmented intelligence
Situations in which machines and humans work together to enhance each other’s efforts when completing tasks.augmented reality
A system that combines real and virtual objects and which runs interactively and in real time.⁷
auditability
Auditability is the ability of an AI system to assess its algorithms, data, and design. Ensuring traceability is key but does not necessarily mean that information about the AI model and intellectual property related to the system is openly available at all times.⁸automation paradox
Situations in which increased automation in tasks leads to a greater need for human oversight and intervention. As machines take over more tasks, the role of humans becomes crucial for monitoring, managing, and stepping in to rectify issues. automation bias
The inclination for humans to favour suggestions from automated systems and to ignore contradictory information. bias
Bias is an anomaly in the output of AI systems, due to the prejudices in the system development process or existing in the training data.⁹big data
Data sets so large and varied they are beyond the capability of traditional data processing.
See also the 4 Vs.blockchain
A decentralised, distributed, digital ledger that is used to record transactions grouped in blocks and that is linked together across many computers. Unlike traditional databases controlled by a central authority, a blockchain operates on a peer-to-peer network, where no single entity has control. This makes it resistant to fraud, as altering one block would require changing all subsequent blocks across the entire network, which is impossible without the consensus of the entire network. This also allows people who do not know each other to trust a shared record of events. See also cryptocurrency.¹⁰botnets
Networks of private computers that are infected with malware and controlled by a ‘botnet agent’ designed to follow the attacker’s instructions without the computer owners’ knowledge. See also distributed-denial-of-service (DDoS) attack.
bring your own device (BYOD)
A policy that allows employees to use their personal devices, such as laptops, smartphones, and tablets, for work purposes instead of using devices provided by the company. buffer overflow attack
A buffer overflow occurs when a system cannot store as much information as it has been sent and consequently may overwrite existing content. A buffer overflow attack occurs when an attacker sends a malicious programme that deliberately overloads the system and attempts to overwrite existing data.business intelligence
Business intelligence (BI) is a set of technological processes for collecting, managing, and analysing organisational data to gain insights that inform business strategies and operations. BI is descriptive, enabling better business decisions that are based on a foundation of current business data. Business analytics (BA) is a subset of BI, with business analytics providing the prescriptive, forward-looking analysis.¹¹

cloud computing
The delivery of on-demand computing resources — everything from applications to data centres — over the internet. It allows individuals and businesses to use these resources without having to manage physical hardware themselves, and only pay for what they use.¹²cobots
Robots working alongside humans (collaborative robots).code-division multiple access (CDMA)
Technology that underpins mobile technology. It has developed rapidly over the last decade and has increased the capability of mobile technology.computer vision
AI systems that enable computers to interpret and understand visual information. By using machine learning and neural networks, computer vision systems can analyse, process, and make decisions based on visual data. Applications include object detection, facial recognition, and autonomous vehicles.

cognitive computing
A technology that aims to mimic human thought processes in complex situations where the answers may be ambiguous. It combines artificial intelligence (AI) and various underlying technologies, such as machine learning, neural networks, natural language processing, and speech and object recognition.controllability
The property of an AI system allowing a human to intervene in its functioning.¹³cryptocurrency
A digital currency that functions as a medium of exchange through a computer network that is not reliant on any central authority, such as a government or bank. Cryptocurrencies typically make use of blockchain technologies.cyber risk
The risk of financial loss, disruption, or damage to the reputation of an organisation caused by issues with the information technology systems they use. The issues could range from errors or unauthorised access to system failure.dark web
A section of the internet that allows anonymity and the ability to obscure the source of information or location of users and website operators. This increased privacy allows criminals to prosper, for example, by selling drugs or stolen payment credentials. Special technology, such as the TOR network, is required to access the dark web.data analytics
The process of collecting, organising, and analysing large sets of data to discover patterns and other information that an organisation can use to inform future decisions. See also big data.data annotation
Linking descriptive information (labels, anchors, or metadata) to data without any change to that data. This practice is essential for training machine learning models, enabling them to accurately interpret and analyse various types of data, including text, images, audio, and video.¹⁴data augmentation
A technique used in machine learning and artificial intelligence to increase the diversity of data available for training models without collecting new data. It involves applying various transformations to existing data to create modified versions of the data. This helps improve the robustness and generalisation ability of models by exposing them to a wider range of variations during training.data lake
A data lake is like a vast reservoir that ingests and stores large volumes of raw data in its original form and allows for big data analysis and machine learning.¹⁵data poisoning
Interference in the AI system that affects the training set, causing the system to learn something it should not learn. By altering the behaviour of the model, data poisoning leads to incorrect output.¹⁶data protection
The process of safeguarding important information from corruption, compromise, or loss.data visualisation
A process that allows large volumes of complex data to be displayed in a visually appealing and accessible way that facilitates the understanding and use of the underlying data.data warehouse
An enterprise system used for the analysis and reporting of structured and semi-structured data from multiple sources. The warehouse will store historic and current data. Unlike a data lake, which stores raw data in its original form, a data warehouse is a repository for processed data, designed for querying and analysis.decision support system (DSS)
An information system that aids managers in making decisions. The system predicts the consequences of possible scenarios; managers then use their judgement to make the final decision.deepfake
Realistic but digitally manipulated video, audio, or other material, generated by using AI. Although the output is convincingly altered, it misrepresents what has been said or done. In other words, it is deeply misleading fake content.

deep learning
A specialised subset of machine learning that uses multi-layered neural networks to analyse complex data and simulate human decision-making. Deep learning algorithms can label and categorise information and identify complex patterns in data. Deep learning enables AI systems to learn continuously and improve the quality and accuracy of results by evaluating the correctness of decisions. Deep learning models can perform unsupervised learning and extract the characteristics, features, and relationships they need to make accurate outputs from raw, unstructured data.¹⁷denial-of-service (DoS) attack
A cyberattack that overwhelms a system’s resources so that it cannot respond to service requests. See also distributed-denial-of service (DDoS) attack.digital asset management (DAM) system
A system designed to coordinate the digital assets of a business, ensuring they are held centrally in an accessible, secure and logically designed repository.

digital assets
Assets held by a business in digital form that do not have physical substance. This typically includes images, animations, audio, video, and PDF files.disruptive technology
An innovation that significantly alters the way that consumers, industries, or businesses operate.distributed-denial-of service (DDoS) attack
A cyberattack that overwhelms a system’s resources and is launched from a large number of other host machines that have already been infected by malicious software controlled by the attacker. See also denial-of-service (DoS) attack. encryption
Encryption is when text information is disguised for security reasons. By using a unique identifier (hash code) that retrieves the piece of data needed, one can access it. Encrypted results are unintelligible when you do not have the encryption key.¹⁸

edge computing
The practice of processing data closer to the source of generation rather than relying on centralised data centres.

enterprise performance management (EPM)
The process of monitoring performance across the enterprise with the goal of improving business performance. An EPM system integrates and analyses data from many sources including e-commerce systems, front-office and back-office applications, data warehouses, and external data sources. Advanced EPM systems can support many performance methodologies such as the balanced scorecard.¹⁹enterprise resource planning (ERP) system
A single system that integrates information from many aspects of operations (for example, manufacturing, inventory management, invoicing, and management accounting) and support functions (such as human resources management and marketing) into one system.ethical hacker
An ethical hacker is someone with the technical knowhow to hack systems and who, with the permission of the company, tries to hack the system to identify weaknesses. There is no malicious intent behind an ethical hacker’s actions.
Also see hacking.executive information system (EIS)
An information system that gives senior management access to internal and external information. Information is presented in a flexible, user-friendly, summarised form with the option to ‘drill down’ to a greater level of detail.expert system
Expert systems hold specialist knowledge, for example, on law and taxation, and allow non-expert users to interact with the system to receive information, advice, and recommended solutions. Can be used at all levels of management.explainability/intelligibility
Explainability is about methods and techniques applied so that AI-generated results are traceable. Intelligibility is about the ability to explain the AI system’s operation in nontechnical terms. AI systems should not be a ‘black box’, where humans cannot explain why the system arrived at a particular output.²⁰
extract, transform, and load (ETL)
The process that blends data from multiple sources into a destination system, such as a data warehouse. Extraction is the process of harvesting data from multiple sources. Transformation takes the extracted data and changes it into a format suitable for the destination database and ultimate intended use. Loading is when the newly cleaned and prepared data is uploaded into the destination.

extranet
An extension of an intranet to include suppliers, customers, and other business partners.

foundation models
Large-scale neural networks trained on vast amounts of data that serve as a base for a multitude of applications. Instead of training a model from scratch for each specific task, you can take a pre-trained foundation model and adapt it for a particular application. General Data Protection Regulation (GDPR)
European Union legislation regulating the collection, processing, and protection of personal data. Similar principles have been adopted in UK laws.general purpose AI
Large AI models trained on a vast quantity of data that can be used to perform a wide range of tasks in a wide variety of fields with minimal adaptation.²¹generative AI
Generative AI is a type of artificial intelligence that creates new content, like text, images, or audio, based on the data it has been trained on. It can produce outputs that resemble human creativity. generative AI (gen AI) chatbot
AI that provides information or assistance to the customer by simulating human conversation.²²
Applications like ChatGPT, Gemini, Copilot, and others use information databases, deep learning, machine learning, and natural language processing to understand the patterns in human conversation. First, the programme analyses the input to recognise intent and key words. Then, with the help of AI algorithms, it categorises the query based on previous interactions and user preferences that improve the relevance. Next, the system selects appropriate responses from the chatbot knowledge base, while also updating it with information from the current conversation. Finally, the response is generated.
groupware
Technology designed to help people collaborate. Examples include email, file-sharing technology, and video/audio conferencing.

hacking
The act of gaining unauthorised access to computer systems, networks, or data.

Hadoop software
Hadoop is an open-source framework based on Java that manages the storage and processing of large amounts of data for applications. Hadoop uses distributed storage and parallel processing to handle big data and analytics jobs, breaking workloads down into smaller workloads that can be run at the same time.²³

hallucination
An AI output that is inaccurate, completely misleading, and fabricated but sounds plausible.

human oversight
Human oversight, or integration of human expertise, in AI is crucial for ensuring ethical, accountable, and effective use of AI technologies. Integrating human expertise with AI systems is essential for a responsible and sustainable future.

This may happen by intervention in each decision cycle (human-in-the-loop) or by monitoring the design and operation of the system (human-on-the-loop). Human-in-command mechanism involves the ability to decide whether and how to use the system or override decisions taken by it.²⁴imputation
Replacing missing data with estimated or modelled data.²⁵

Information to impact infinity loop

‘information to impact’ infinity loop
Also see the 5-As framework. The finance function’s basic activities form the ‘information to impact’ infinity loop. The loop consists of assembling information, analysing for insights, advising to influence, applying for impact, and assessing to assemble.

internet of things (IoT)
A network of smart devices with inbuilt software and connectivity to the internet allowing them to constantly monitor and exchange data (such as the fitness trackers in smart watches or smart thermostats that automatically adjust temperature to your preferences).

intranet
A private network within an organisation allowing the sharing of information and resources with employees.ISO/IEC 27001
Internationally recognised standard that provides companies with guidance for establishing, implementing, maintaining, and continually improving an information security management system.²⁶

keylogger
Type of spyware where every keystroke typed by the victim is recorded and forwarded to the hacker, allowing them to record usernames and passwords. knowledge management
The process for the acquisition, sharing, retention, and utilisation of knowledge.knowledge management system
Any type of IT that helps capture, store, retrieve, and use knowledge to enhance the knowledge management process label
A marker assigned to raw data (images, text files, videos, etc) that describes it so that an AI system can learn from the data.²⁷large language models (LLM)
Large language models (LLMs) are a type of foundation model trained on vast datasets. Such systems understand and can reproduce natural human conversations and can perform a variety of tasks such as generating text, translating languages, summarising information, decision-making, and problem-solving.²⁸

lemmatisation
Lemmatisation is a text preprocessing technique in natural language processing (NLP) that involves reducing words to their base or dictionary form, known as the lemma. Unlike stemming, which simply cuts off word endings, lemmatisation considers the context and the morphological analysis of words to ensure that the base form is meaningful and accurate.²⁹

low displacement potential
The ability of AI systems to only partially perform certain occupations that require reasoning in novel conditions, interpersonal skills, or complex manual jobs even though some tasks related to those roles can be automated.³⁰

machine learning
A subset of AI where computer code is built to mimic how the human brain works. Essentially, machine learning algorithms detect patterns and learn how to make predictions and recommendations rather than simply following explicit programming instructions.

malvertising
Online advertisements that have malware written into their code. It can involve hiding the malicious code in legitimate online advertising networks and web pages. The code may direct the victim to a malicious site where the malware can be installed, or it may directly infect the victim’s computer when they visit the page.

malware
Malicious software, regardless of the intended purpose. It can do any number of things, ranging from stealing credentials, other information, or money, to general widespread destruction, or denial of service.man-in-the-middle attacks (MitM)
Cyberattack where the attacker secretly and independently makes connections between two parties and passes messages between them. The aim is to make them think they are communicating directly with each other while the attacker is controlling the conversation.

material requirement planning (MRP)
A computerised system for planning the requirements for raw materials, work-in-progress, and finished items. See also manufacturing resource planning (MRPII).manufacturing resource planning (MRPII)
Developed from material requirement planning (MRP), MRPII also includes production planning, machine capacity scheduling, demand forecasting and analysis, quality tracking tools, employee attendance, and productivity tracking.metadata
Data characteristics that facilitate the manipulation and analysis of a certain data set, aside from its content. Metadata provides descriptors about data ownership, access paths, and more.³¹mixed reality
A mixed reality environment is one in which real-world and virtual-world objects are presented together within a single display.³²mobile technology
Technology that goes where the user goes. It consists of portable two-way communications devices, computing devices, and the networking technology.

moral agent
AI system that can recognise and apply the moral aspects of a situation and make decisions accordingly.³³narrow AI
AI systems that solve a specific problem and, hence, operate within a limited, pre-defined range of functions.³⁴natural language
Language used in a community of people that can be expressed in speech, text, sign language, and so on.³⁵

no displacement potential
The idea that certain jobs or roles are less likely to be replaced by AI and automation because they often require uniquely human skills such as creativity, emotional intelligence, complex problem-solving, and interpersonal communication.Payment Card Industry Data Security Standards (PCI DSS)
Standards enacted to ensure the ways businesses process card payments are secure, protect sensitive cardholder data, and help reduce card fraud. This is done by setting tight controls over the storage, transmission, and processing of the cardholder data that businesses handle.

phishing
The use of fraudulent messages to try to steal sensitive information such as passwords or credit card numbers, or to install malware onto a user’s computer. See also spear phishing.polymorphic malware
A type of malware that avoids being identified by systems and networks by constantly changing its identifiable features.predictive analytics
Data analysis that draws on historical data, both internally and externally sourced, to anticipate future outcomes and trends.³⁶process automation
The use of technology to automate complex business processes, either in their entirety or specific elements, to improve consistency, quality, and speed, while also achieving cost savings.

ransomware
A type of malicious software designed to encrypt or lock a business’s data, information, or entire computer system, preventing access until a specified ransom is paid.retrieval-augmented generation (RAG)
A technique used in gen AI models to enhance their capabilities by incorporating external information capabilities of LLMs with an information-retrieval system. When a user asks a question, the model retrieves relevant information from external sources (like databases, documents, or the web) and incorporates this information into its response, using it to supplement its preexisting training data.RAT (remote access trojan)
A type of trojan malware that is remotely controlled by the attacker, enabling the hacker to perform various malicious activities. See also trojan.reinforcement learning
This type of machine learning relies on providing a set of rules and constraints and letting the algorithm learn how to achieve its goals. Humans define the state, the desired goal, allowed actions, and constraints. The algorithm figures out how to achieve the goal by trying different combinations of allowed actions and is rewarded or sanctioned, depending on whether the decision was a good one.

responsible AI
The practice of designing, developing, and deploying artificial intelligence systems in a way that ensures they are ethical, transparent, and accountable. Responsible AI emphasises the importance of fairness, privacy, and security, aiming to minimise bias and prevent harm.

robot
A programmed machine that performs intended tasks in the physical or digital realm, by sensing its environment and responding to a software control system.³⁷
robotic process automation (RPA)
The use of robotics to automate routine activities. Many believe that the automation of routine and repetitive tasks is a critical part of the digital transformation in finance and other functions.

robotics
Designing, constructing, and operating robots — machines that can perform tasks on their own or with some help.

robustness
AI is robust when it can effectively handle exceptional conditions like abnormal inputs or adversarial attacks.

semi-supervised machine learning
Machine learning that combines elements of both supervised and unsupervised learning by making use of both labelled and unlabelled data during training models.³⁸

sentiment analysis
Task of computationally determining the emotional tone behind a piece of text, speech, or image to determine feelings and attitudes on a scale from positive to neutral to negative. Also known as opinion mining or emotion AI.³⁹

social engineering
The manipulation of people into compromising their security. It aims to make them perform specific actions or reveal confidential information.⁴⁰speech synthesis/text-to-speech (TTS)
Generation of artificial speech by converting written text from digital sources to audio output.⁴¹soft computing
Soft computing is a branch of computing that offers approximate solutions to complex problems, rather than exact solutions. It is applied in areas where traditional computing methods are limited, such as pattern recognition, data mining, and control systems. By employing methodologies that replicate human reasoning and decision-making, it allows for flexibility and adaptability in problem-solving.
spear phishing
When a phishing attempt targets a specific user, rather than a blanket communication sent to many people. Spear phishing usually involves carrying out research into the specific person and their role and interests. This allows the attacker to modify the communication to be more appealing or relevant to the victim, increasing the likelihood of its success.spyware
Malware designed to spy on the victim’s systems without being detected and gather information to send to the hacker. See also keylogger. stemming
A text preprocessing technique in natural language processing (NLP) that involves reducing words to their root form to help with text analysis, information retrieval, or sentiment analysis.⁴²SSL certificate
An SSL certificate verifies the sender’s identity, enables encrypted transmission of confidential information, and ensures the receiver knows if the data has been tampered with, much like virtual tamper-proof tape. strong AI
AI possessing diverse capabilities across unrelated tasks; it can acquire new skills or face new challenges, learn from experience, and perform at a human level of intelligence.

super AI (conscious AI)
AI with human-level consciousness and self-awareness. Its development is currently considered impossible, as we are unable to adequately define consciousness.

superintelligence
A hypothetical agent that surpasses human cognitive abilities across all domains, including science, creativity, and social intelligence.⁴³supervised learning
A type of machine learning in which an algorithm is trained on human-labelled data and is provided with samples to become precise in classifying new data. symbolic AI
Also known as good old-fashioned AI (GOFAI).
An approach to artificial intelligence that relies on the manipulation of symbols and the use of explicit rules to represent knowledge and solve problems. It is particularly effective in domains where problems can be clearly defined and rules can be explicitly stated. In contrast, modern AI approaches, like neural networks and machine learning, rely more on data-driven methods and statistical learning.

test data
In the context of machine learning, test data is used to assess the performance of trained and validated AI systems and to draw conclusion about their reliability and generalization capabilities.

threat actor
A term for an individual or group that, either intentionally or unintentionally, conducts malicious activities against an organisation. It can include negligence or mistakes by a person or a group of people who do not have malicious intent.

TLS certificate
A TLS (transport layer security) certificate is an updated version of an SSL certificate, which aims to speed up the verification process. See also SSL certificate.

tokenization
In the context of natural language processing (NLP), this is the process of splitting text into smaller units so that each one represents a meaningful unit of the larger whole. This allows NLP tasks like translation, classification, or sentiment analysis by enabling machines to recognise patterns, analyse structures, and derive meaning.traceability
The ability to trace data through all stages of input, sampling, labelling, processing, and decision-making.⁴⁴training data
Samples of the input data used to train the machine learning model.⁴⁵transaction processing system (TPS)
An information system that helps operational managers make decisions. It records all the daily transactions (for example, payroll or purchases) of an organisation and summarises them, so they can be routinely reported.

transparency
AI is considered transparent when it provides clear and relevant information about its design and development processes.trojan
Software that pretends to be useful while secretly releasing malware into the system, usually with the capability to be controlled by the attacker from a different location. Once on the system, it can then prevent access to the system (ransomware), infect the system, damage or destroy files, or act as spyware.trustworthy AI
Refers to the reliability and integrity of the AI system itself, as well as all processes and actors involved in the AI life cycle. Trustworthy AI encompasses three key aspects:

Adherence to laws and regulations.

Respect for ethical principles and values.

Ensuring the system is robust from both technical and social perspectives, meaning it does not cause harm.⁴⁶

Turing test
Named after the British mathematician and computer scientist Alan Turing. The test, originally called ‘the imitation game’, measures a machine’s ability to exhibit intelligent behaviour indistinguishable from that of a human. It is believed that when this is achieved, human and machine intelligence will have reached equivalenceunstructured data
Unstructured data lacks a predefined data model and isn’t organised systematically. Unlike structured data, organised in tables and databases, it includes varied formats like text, images, audio, and video. Advanced techniques such as NLP, computer vision, and machine learning are required to derive insights from unstructured data.

unsupervised learning
A type of machine learning where the algorithm is provided with unlabelled data and tasked with identifying patterns and drawing inferences on its own. This approach is particularly useful for clustering data, where data points are grouped based on their similarity to neighbouring points and their dissimilarity to others. Once the data is clustered, various techniques can be employed to explore the data and uncover patterns. use case
A scenario in which AI can be used (for instance, self-driving cars).⁴⁷validation data
A sample of data held back from training an AI system that is used to give an estimate of model skill while adjusting the model’s hyperparameters (configurations set before the training starts).⁴⁸

virtual private network (VPN)
The extension of a private network across a public network, enabling users to send and receive data across shared or public networks as if their computing devices were directly connected to the private network. Often used by organisations whose staff work remotely so that staff can access shared data drives or the organisation’s intranet.

virus
A type of malicious software designed to endlessly replicate itself and infect programs and files to damage or destroy data. See also worm virus.

weak AI (narrow AI)
AI tailored for a specific domain. Makes decisions based on programmed algorithms and training data (for instance, language translations, virtual assistants).weaponised document
A document that is downloaded from a source (email, website, or shared drive) that contains some code, a link, or even a video that, once activated, releases malware onto a system or network.workflow of the AI model
Iterative phases needed to build the AI model, including data collection, model development, training, accuracy evaluation, usage, maintenance, versioning, and so on.⁴⁹worm virus
A type of virus that can spread across networks to infect other devices.

CIMA wants to hear from you once you’ve reviewed the glossary!