x®

ARTIFICIAL INTELLIGENCE GLOSSARY

Absolute Intelligence: A theoretical form of intelligence representing maximal cognitive, emotional and computational capabilities, beyond human or current artificial capacities. It implies an intelligence that has surpassed all known limitations of human understanding and processing power.
Active Learning: A machine learning approach where models iteratively select the most informative data points for training to improve performance efficiently. This technique is particularly useful in situations where labeled data is scarce, as it allows the model to learn more effectively from fewer examples.
Advanced Intelligence: Intelligence systems capable of superior reasoning, adaptation and multi-domain problem-solving. These systems typically exhibit cognitive flexibility and sophisticated problem-solving abilities that far exceed those of basic artificial intelligence.
Agentic Intelligence: Artificial intelligence exhibiting autonomous, goal-directed behavior, demonstrating initiative, planning and self-modulation in dynamic environments. These systems can independently pursue objectives and adapt their behavior based on the environment, resembling human-like decision-making processes.
Algorithmic Bias: Systematic deviations in artificial intelligence outputs caused by flawed data, assumptions, or algorithms. This bias can perpetuate inequalities or inaccuracies, especially when artificial intelligence is trained on biased data sets or when the algorithms fail to account for important contextual factors.
Alternative Intelligence: This refers to non-human forms of intelligence, often characterised by novel or unconventional methods of problem-solving, learning, or reasoning, which differ from traditional human cognitive processes.
Ambient Intelligence: Artificial intelligence systems embedded into the environment, allowing them to respond intelligently to human actions and context. These systems are designed to seamlessly integrate into everyday life, such as in smart homes or smart cities, reacting to changes in the environment to improve user experience.
Analogical Reasoning: The cognitive or computational process of solving new problems by relating them to previously encountered situations. By drawing parallels between similar scenarios, systems can apply existing knowledge to novel tasks, often used in problem-solving and decision-making.
Anomaly Detection: Artificial intelligence techniques identifying unusual patterns or deviations from expected behavior in data. This method is commonly used in fraud detection, network security and quality control, helping to identify outliers or errors that may indicate a problem.
Artificial Consciousness: A hypothetical artificial intelligence capable of self-awareness, subjective experience and reflective thought. Unlike typical artificial intelligence, which executes tasks based on predefined programming, artificial consciousness would possess a sense of being and the ability to introspect.
Artificial General Intelligence: Artificial intelligence possessing flexible, human-comparable cognitive abilities across multiple domains. Artificial general intelligence is capable of understanding, learning and applying knowledge in a wide range of contexts, just like a human and can adapt to new tasks without requiring retraining.
Artificial Intelligence: The field of study and practice focused on creating systems capable of performing tasks that normally require human cognition. These tasks include problem-solving, pattern recognition, decision-making and language understanding, among others.
Artificial Neural Networks: Computational models inspired by biological neural networks, used to approximate complex functions and patterns in data. Artificial neural networks consist of interconnected nodes (neurons) arranged in layers and are widely used for tasks like image recognition and natural language processing.
Artificial Superintelligence: Hypothetical intelligence that far surpasses human intelligence across all domains. Artificial superintelligence would not only outperform humans in specific areas but also possess cognitive abilities that allow it to improve and innovate autonomously.
Bionic Brain: A hybrid cognitive system combining biological and artificial neural processing. The bionic brain concept often involves integrating artificial components with human or animal neural networks to enhance cognitive abilities or restore brain functions.
Bionic Intelligence: Intelligence arising from the integration of human-like and artificial cognitive elements. This could involve combining biological brain functions with artificial enhancements to achieve higher cognitive performance or to overcome limitations of the human brain.
Bionic Mind: A singular hybrid cognitive system combining human and artificial cognition. This concept explores the potential for humans to directly interface with artificial systems in ways that enhance or expand mental capabilities.
Bias Mitigation: Techniques to identify and reduce unfair or skewed outcomes in artificial intelligence systems. Bias mitigation is critical in artificial intelligence deployment, especially in sensitive areas like hiring, law enforcement and healthcare, to ensure fairness and avoid reinforcing existing inequalities.
Big Data: Extremely large datasets that require specialised computational techniques to analyse and extract insights. Big data is used in numerous fields, including business, healthcare and social sciences, to uncover patterns, make predictions and support decision-making.
Collective Intelligence: The shared or group intelligence that emerges when multiple entities collaborate, often surpassing individual capabilities. This concept is widely used in artificial intelligence for systems that leverage multiple agents or contributors to solve problems more effectively, such as in crowdsourcing or swarm intelligence.
Cognitive Computing: Artificial intelligence systems simulating human thought processes for reasoning, problem-solving and decision-making. These systems aim to mimic human cognition by interpreting and responding to complex data, often using natural language processing and machine learning.
Cognitive Science: Interdisciplinary study of cognition, intelligence and behavior, informing artificial intelligence research. It combines insights from psychology, neuroscience, computer science, linguistics and philosophy to understand how intelligence works and how to replicate it in machines.
Computational Creativity: Artificial intelligence techniques designed to generate novel ideas, art, or solutions. Computational creativity involves algorithms that can produce music, visual art, poetry and other forms of creative expression that were traditionally considered uniquely human.
Computational Intelligence: Artificial intelligence approaches using heuristic, probabilistic and optimisation methods rather than purely symbolic logic. These methods focus on solving complex, real-world problems by mimicking processes found in nature, such as evolutionary algorithms and neural networks.
Computer Vision: Artificial intelligence techniques enabling machines to interpret and process visual information. Computer vision allows artificial intelligence systems to recognise objects, track movement and understand images or video, which is essential for applications in self-driving cars, facial recognition and medical imaging.
Convolutional Neural Networks: Deep learning architectures effective for image and spatial data analysis. These are particularly well-suited for tasks like image classification, object detection and video analysis, as they are designed to automatically learn spatial hierarchies of features in images.
Contrastive Learning: A machine learning approach where models learn by comparing similarities and differences between data representations. This method is used to improve the model’s ability to identify key features by training it to differentiate between related and unrelated examples.
Clustering: Unsupervised learning technique that groups similar data points together. Clustering is often used in data analysis to identify patterns or segments in large datasets, such as customer segmentation in marketing or anomaly detection in security systems.
Data Augmentation: Techniques used to increase dataset diversity without additional data collection. By applying transformations like rotation, cropping, or flipping to existing data (e.g., images), data augmentation helps improve the generalisation ability of machine learning models.
Data Mining: Extraction of patterns and insights from large datasets using algorithms. Data mining involves using statistical and machine learning techniques to discover hidden relationships, trends and insights in data, often for applications in business and research.
Data Science: Interdisciplinary field combining statistics, AI and domain knowledge. Data science focuses on extracting useful information from large datasets, including data cleaning, analysis, visualisation and predictive modelling.
Decision Support Systems: Tools assisting human decision-making through data analysis. These systems integrate data from various sources, provide recommendations and often use artificial intelligence to simulate possible outcomes to help make more informed decisions.
Deep Learning: A subset of machine learning using multi-layered neural networks to model complex patterns. Deep learning is highly effective in tasks such as image recognition, natural language processing and speech recognition, as it automatically learns feature hierarchies from raw data.
Dimensionality Reduction: Techniques to reduce the number of variables in data. Dimensionality reduction helps to simplify models, reduce computational costs and improve the performance of machine learning algorithms, often using methods like principal component analysis.
Edge Intelligence: Artificial intelligence deployed at the edge of networks for real-time processing. Edge intelligence allows devices to process data locally, reducing the need for cloud-based processing and enabling faster decision-making in real-time applications like IoT and autonomous vehicles.
Enhanced Intelligence: Intelligence amplified through artificial intelligence systems. Enhanced intelligence refers to human cognitive abilities augmented by artificial intelligence tools, enabling people to process and analyse more information, make better decisions and solve problems faster.
Enterprise Intelligence: Artificial intelligence applied to optimise organisational decision-making. Enterprise intelligence involves using artificial intelligence-driven insights to improve business operations, customer relationships and strategic planning, often by analysing large datasets and automating tasks.
Ethical AI: Artificial intelligence aligned with fairness, accountability and transparency. Ethical AI ensures that artificial intelligence systems are developed and deployed responsibly, with attention to avoiding harm, reducing bias and ensuring that decisions are made in a way that respects human rights.
Evolutionary Algorithms: Algorithms inspired by natural selection. These algorithms evolve solutions over time by iterating through generations, selecting the best-performing solutions and combining them to produce new candidates. Evolutionary algorithms are often used for optimisation problems.
Explainable AI: Methods enabling humans to understand artificial intelligence decisions. Explainable AI seeks to make the decision-making process of systems transparent and interpretable, allowing users to trust and verify artificial intelligence outputs, especially in high-stakes areas artificial intelligence like healthcare and finance.
Feature Engineering: The process of transforming raw data into meaningful features that can be used in machine learning models. Feature engineering is crucial for improving the performance of a model, as it helps extract the most relevant information from the data for better predictive accuracy.
Fine-Tuning: The process of adapting a pre-trained model to a specific task or dataset. Fine-tuning is often done by training the model on a smaller, specialised dataset after it has been initially trained on a larger, more general one. This allows the model to adapt its general knowledge to a particular domain.
Frontier Intelligence: Cutting-edge artificial intelligence approaches that push the boundaries of what is currently possible. Frontier intelligence involves research and development of advanced artificial intelligence techniques that are at the forefront of innovation, including next-generation algorithms, novel architectures and emerging applications.
Fuzzy Logic: A reasoning method that allows for handling uncertainty and imprecision. Unlike traditional Boolean logic (which deals with true/false values), fuzzy logic uses degrees of truth, allowing artificial intelligence systems to make decisions based on vague or incomplete information. It's often used in control systems and decision-making.
Generative AI: Artificial intelligence systems capable of producing novel content, such as images, music, or text. These systems learn patterns from training data and then generate new, similar content, enabling tasks like automated content creation, design and product generation.
Generative Adversarial Networks: A type of machine learning architecture consisting of two competing neural networks: a generator and a discriminator. The generator creates new data, while the discriminator evaluates its authenticity. Through this adversarial process, generative adversarial networks can generate high-quality images, videos and other types of content.
General Intelligence: Intelligence capable of performing a wide range of cognitive tasks across various domains. Unlike weak artificial intelligence (which excels in specific tasks), general intelligence can adapt, learn and apply knowledge flexibly in new situations, similar to human intelligence.
Genius: Exceptional problem-solving or creativity beyond the typical human or artificial capabilities. In the context of artificial intelligence, genius could refer to an artificial intelligence system that demonstrates extraordinary creativity, insight, or innovation, often leading to novel solutions in complex scenarios.
Guardrails: Safety mechanisms ensuring artificial intelligence behaves within acceptable limits. Guardrails are implemented to prevent artificial intelligence systems from making harmful or undesirable decisions. These can include ethical constraints, safety checks, or performance boundaries to ensure that artificial intelligence operates safely and responsibly.
Hallucinations: Incorrect or fabricated outputs generated by artificial intelligence systems. In natural language processing or image generation, hallucinations occur when a model generates information that seems plausible but is factually inaccurate or made up, often because the model has learned from noisy or incomplete data.
Heuristics: Problem-solving strategies that use shortcuts to find solutions more quickly, though they do not always guarantee optimal results. Heuristics are used in artificial intelligence to make decisions more efficiently, particularly when time or computational resources are limited. They are often applied in decision-making and search algorithms.
Hyperintelligence: Intelligence that exceeds current human or artificial levels. Hyperintelligence would involve cognitive abilities that are far beyond the capacity of any existing system, capable of solving problems, making decisions and understanding concepts in ways that humans cannot fully grasp.
Intelligence: The capacity to learn, reason and solve problems. It refers to the ability to adapt to new situations, understand complex concepts and apply knowledge to achieve goals. In artificial intelligence, intelligence is about creating systems that can simulate or replicate these human-like cognitive abilities.
Intelligent Agents: Systems that perceive their environment, make decisions and take actions autonomously. Intelligent agents can function in dynamic environments, adapt to changes and pursue specific goals or objectives without needing constant human intervention.
Interpretability: The degree to which humans can understand the decisions made by an artificial intelligence system. Interpretability is crucial for building trust in artificial intelligence, especially in high-stakes fields like healthcare or criminal justice, where users need to know how decisions are made to ensure fairness and accountability.
Intuitive AI: Artificial intelligence that simulates human-like intuition. Intuitive artificial intelligence systems can make decisions quickly, often relying on heuristics or prior experience rather than detailed, step-by-step reasoning. These systems aim to replicate the instinctive judgment humans might make in complex or ambiguous situations.
Joint Attention: The shared focus between artificial intelligence and humans on an object or event. Joint attention is crucial for social interaction and learning and in artificial intelligence, it refers to systems that can recognise and engage with human attention, such as in collaborative robotics or conversational agents
Knowledge Graphs: Structured representations of entities (such as people, places, or concepts) and their relationships. Knowledge graphs help artificial intelligence systems understand the context and relationships between different pieces of information, enabling more effective reasoning, search and decision-making.
Large Language Models: Artificial intelligence models trained on vast amounts of text data to understand and generate human language. These models can perform a wide range of language-related tasks, including text generation, translation, summarisation and question answering.
Latency: The time delay between initiating a request and receiving a response. In artificial intelligence systems, latency can be a critical factor in real-time applications, such as autonomous driving or interactive artificial intelligence, where quick decision-making and responsiveness are essential.
Latent Space: An abstract, often high-dimensional representation of the underlying features or patterns in data. In deep learning, latent spaces are used to transform raw data (like images or text) into more structured forms, which can then be manipulated or analysed by models.
Linear Regression: A statistical method for modelling the relationship between a dependent variable and one or more independent variables. Linear regression is widely used in machine learning for predicting continuous outcomes, such as predicting house prices based on factors like size and location.
Model Context Protocol: A framework or set of rules designed to manage and maintain context within artificial intelligence models. It ensures that models can accurately interpret and respond to inputs in a way that is consistent with previous interactions or the overall task at hand. This is particularly useful for tasks requiring long-term memory or complex decision-making processes.
Machine Learning: A subset of artificial intelligence that focuses on developing algorithms and models that allow computers to learn patterns and make decisions based on data, without being explicitly programmed. Machine learning techniques include supervised, unsupervised and reinforcement learning, among others.
Meta-Learning: Often referred to as "learning to learn," meta-learning involves creating models that can adapt and improve their learning process over time. Meta-learning helps AI systems transfer knowledge from one task to another and improve learning efficiency, even with limited data.
Multimodal AI: Artificial intelligence systems that integrate and process multiple types of data (e.g., text, images, audio, video, etc.) simultaneously. Multimodal AI can perform more complex tasks, such as understanding a video with both visual and spoken content or responding to a voice command with both spoken and visual feedback.
Model Explainability: Techniques and methods used to make machine learning models more transparent, helping users understand how and why certain decisions were made. This is crucial for trust, accountability and ensuring that artificial intelligence behaves predictably and ethically, especially in critical applications like finance and healthcare.
Natural Language Generation: A subfield of natural language processing that involves creating human-like text from structured data. These models can be used to generate reports, summaries and creative content automatically, based on the input data or user prompt.
Natural Language Processing: A field of artificial intelligence that enables computers to understand, interpret and generate human language in a way that is both meaningful and contextually relevant.
Natural Language Understanding: The ability of artificial intelligence systems to understand and interpret human language in a way that conveys meaning. This allows machines to recognise the intent behind user inputs, which is crucial for applications like chatbots, virtual assistants and sentiment analysis.
Neural Networks: Artificial intelligence models inspired by the structure and functioning of the human brain, designed to recognise patterns and solve complex problems. Neural networks consist of layers of interconnected "neurons" (nodes) that process data in a hierarchical manner. Deep neural networks, with multiple layers, are particularly powerful for tasks like image and speech recognition.
Neuro-Symbolic AI: A hybrid approach that combines the strengths of neural networks (which excel at pattern recognition) and symbolic reasoning (which uses logic and structured knowledge). Neuro-symbolic AI aims to create systems that are both data-driven and capable of abstract reasoning, improving performance on tasks requiring common-sense understanding.
Overfitting: A common problem in machine learning where a model becomes too tailored to the training data, capturing noise and irrelevant details, leading to poor generalisation to new, unseen data. Overfitting reduces the model’s ability to perform well on real-world tasks and is mitigated by techniques like cross-validation and regularisation.
Optimisation: The process of adjusting the parameters of a machine learning model to maximise or minimise a particular objective, such as reducing error or improving predictive accuracy. Optimisation algorithms, like gradient descent, are used to find the best possible solution within a model’s parameter space.
Quantum Intelligence: A nascent field that merges artificial intelligence with quantum computing. Quantum computing has the potential to solve certain problems exponentially faster than classical computers and combining it with artificial intelligence could lead to breakthroughs in areas such as cryptography, optimisation and machine learning. Quantum intelligence could revolutionise tasks that are computationally intractable for traditional systems.
Recurrent Neural Networks: A type of neural network designed for sequential data, such as time series or text. Recurrent neural networks have "memory" by incorporating loops in their architecture, which allows them to use information from previous inputs to make better predictions about future data. They are often used in applications like speech recognition and language modelling.
Reinforcement Learning: A type of machine learning where an agent learns to make decisions by interacting with an environment and receiving feedback in the form of rewards or penalties. Over time, the agent improves its actions to maximise the cumulative reward. Reinforcement learning is widely used in robotics, game artificial intelligence and autonomous vehicles.
Responsible AI: The development and deployment of artificial intelligence technologies that are ethical, transparent and aligned with societal values. Responsible AI involves considering fairness, accountability and safety throughout the lifecycle of artificial intelligence systems, from design and deployment to monitoring and governance.
Robotics: The field of engineering and artificial intelligence concerned with designing and creating machines that can perform tasks autonomously or semi-autonomously. Artificial intelligence plays a key role in making robots intelligent, enabling them to perceive their environment, make decisions and interact with humans and objects.
Self-Supervised Learning: A type of unsupervised learning where the model learns to predict part of the data from the rest, generating its own "labels" or supervision signals. This approach is particularly useful in scenarios where labeled data is scarce or expensive to obtain.
Semi-Supervised Learning: Combines labeled and unlabelled data to train a machine learning model. It is particularly helpful in situations where acquiring labeled data is expensive or time-consuming, allowing models to benefit from large amounts of unlabelled data while still incorporating a smaller set of labeled examples.
Superhuman Intelligence: Intelligence that exceeds human cognitive abilities across all domains. This term is often used in discussions about the potential future development of artificial superintelligence, where artificial intelligence could outperform humans in virtually every intellectual task.
Superintelligence: Refers to a form of intelligence that surpasses human capabilities across every domain, including reasoning, creativity, social intelligence and more. Superintelligent systems would have the ability to improve themselves rapidly, leading to a significant and unpredictable shift in technological and societal dynamics.
Superintelligent: Describes an artificial intelligence system that possesses superhuman intelligence. Superintelligent artificial intelligence could outthink and outperform humans in every aspect and is often considered the ultimate goal in the development of artificial general intelligence and artificial superintelligence
Supervised Learning: A type of machine learning where the model is trained on labeled data. The input data is paired with the correct output (or label) and the model learns to map inputs to outputs. This is one of the most common methods used in machine learning, used for tasks like classification, regression and prediction.
Swarm Intelligence: A type of artificial intelligence inspired by the collective behavior of social organisms like ants, bees, or birds. Swarm intelligence systems involve multiple simple agents that work together to solve problems, often resulting in emergent behavior that is more efficient than any individual agent’s actions.
Synthetic Intelligence: Refers to artificial intelligence systems created through artificial means, rather than emerging from biological systems. Synthetic intelligence contrasts with natural intelligence (such as human or animal cognition) and encompasses the entire field of artificial intelligence development.
Strong AI: A form of artificial intelligence that not only simulates human intelligence but also possesses actual understanding and consciousness. Unlike weak AI, which is designed to perform specific tasks, strong AI would have the ability to think, reason and understand at a human level.
Transfer Learning: A machine learning technique where knowledge gained from solving one problem is applied to a different but related problem. Transfer learning is especially useful when there is limited data for the target task, allowing models to leverage pre-trained knowledge to achieve better results with less data.
Turing Test: A test of a machine’s ability to exhibit intelligent behavior indistinguishable from that of a human. The test, proposed by Alan Turing in 1950, involves a human evaluator interacting with both a machine and a human through text and the machine passes the test if the evaluator cannot reliably tell which is which.
True Intelligence: The idealised concept of intelligence that encompasses both the depth and flexibility seen in human cognition, combined with the perfection of computational processes. True intelligence would integrate emotional, cognitive and sensory faculties seamlessly, making decisions that are not only logical but contextually and morally aware.
Trustworthy AI: Artificial intelligence systems that are reliable, ethical and responsible. Trustworthy AI focuses on transparency, accountability and fairness, ensuring that artificial intelligence systems make decisions that align with societal values and can be trusted by users and stakeholders.
Unsupervised Learning: A type of machine learning where the model is trained on data without explicit labels. The goal is for the model to identify underlying patterns or structures in the data, such as clustering similar items together or discovering hidden relationships.
Universal Intelligence: A theoretical form of intelligence that is applicable across all domains and capable of solving any problem. Universal intelligence would be able to adapt to new environments and tasks without requiring specialised programming or training, much like human intelligence.
Weak AI: Systems that are designed to perform specific, well-defined tasks. These systems are highly effective in their domain but lack the broad cognitive abilities or adaptability of humans.

This website is owned and operated by X, a trading name and registered trade mark of
GENERAL INTELLIGENCE PLC, a company registered in Scotland with company number: SC003234