AI Dictionary
ELOQUENCE’s AI Dictionary for Beginners: Need-to-Know Terms and Definitions
Artificial Intelligence (AI) has been thrust into the spotlight, bringing with it a host of phrases, acronyms, and concepts that, until recently, were hardly used outside of computer science. It’s fast becoming essential to have an understanding of these terms. If this new lexicon is overwhelming you, don’t worry — we’ve got your back. Here’s your pocket dictionary of the most common, need-to-know terms in artificial intelligence.
50+ AI terms and phrases to know:
Artificial Intelligence (AI)
The simulation of human intelligence processes by machines, especially computer systems.
Machine Learning (ML)
Subset of AI focusing on algorithms that enable computers to learn from and make predictions or decisions based on data.
Neural Network
Computational models inspired by the human brain, used in machine learning to recognize patterns and make decisions.
CNNs
Convolutional Neural Networks (CNNs) are specialized neural networks adept at processing data with a grid-like topology, like images, using convolutional layers; key in image recognition tasks.
RNNs
Recurrent neural networks (RNNs) are designed to recognize patterns in sequences of data, such as text or time series, by using their internal state (memory) to process sequences.
GANs
Generative Adversarial Networks (GANs): a framework of two neural networks contesting with each other in a game (generative and discriminative); widely used in image generation, style transfer, and more.
Deep Learning
A subset of Machine Learning using neural networks with multiple layers to model complex patterns in data. It’s fundamental to LLMs.
Generative AI
Generative AI (GenAI) refers to artificial intelligence systems that can generate new content or data that is similar but not identical to the data they were trained on, often used for creating images, videos, sounds, or text in a variety of domains. Not to be confused with AGI.
Foundation Models
A foundation model is a type of large-scale AI model that is pre-trained on a vast amount of data across various domains and can later be fine-tuned for specific tasks. These models serve as a base for further specialization. Training foundation models can cost billions of $US.
LLMs
Large Language Models (LLMs) are advanced AI models that process and generate human-like text, relying on ML and DL techniques.
SLMs
A small language model is a scaled-down language model designed for language processing tasks. It’s less complex and powerful than lfull LLMs, offering reduced computational requirements and often faster response times.
Diffusion Models
A small language model is a scaled-down language model designed for language processing tasks. It’s less complex and powerful than lfull LLMs, offering reduced computational requirements and often faster response times.
Natural Language Processing (NLP)
A branch of AI that focuses on enabling machines to understand, interpret, and respond to human languages, used in applications such as translation, sentiment analysis, and chatbots.
Computer Vision
A field of AI focused on enabling machines to interpret and understand the visual world from digital images and videos, used in object detection, image classification, etc.
Supervised Learning
A training method where models learn from labeled data, using input-output pairs to predict outcomes. Common in classification and regression tasks.
Self-Supervised Learning
A form of unsupervised learning where the model generates its own labels from the input data, often used in language model training.
Semi-Supervised Learning
Combines both labeled and unlabeled data for training. Useful when acquiring a fully labeled dataset is costly or impractical.
Unsupervised Learning
Involves training models on data without labels. The model detects patterns and structures in the data, often used for clustering and association.
Reinforcement Learning
An approach where models learn to make decisions by performing actions and receiving feedback, often used in gaming, navigation, and real-time decisions.
Algorithm
A set of rules or steps followed by a computer to perform a task or solve a problem.
Big Data
Extremely large data sets that may be analyzed computationally to reveal patterns, trends, and associations.
Data Mining
The practice of examining large databases to generate new information.
Feature Extraction
The process of transforming raw data into a set of features that can be used in ML.
Overfitting
A modeling error in ML where a model is too closely fit to a limited set of data points.
Underfitting
A modeling error in ML where a model is too simple to capture the underlying pattern of the data.
Bias
A systematic error introduced into sampling or testing by selecting or encouraging one outcome or answer over others.
Variance
The degree of spread in a set of data values.
Hyperparameter
A parameter whose value is set before the learning process begins.
Training Data
A dataset used to train a model.
Validation Data
A dataset used to provide an unbiased evaluation of a model fit during the training phase.
Training Data
The dataset used for training AI models. The quality and quantity of learning data are crucial for the accuracy and effectiveness of the trained model.
Confusion Matrix
A table used to describe the performance of a classification model.
Accuracy
The ratio of correctly predicted instances to the total instances.
Precision
The ratio of correctly predicted positive observations to the total predicted positives.
Recall (Sensitivity)
The ratio of correctly predicted positive observations to all observations in the actual class.
F1 Score
A measure of a test’s accuracy that considers both the precision and the recall.
ROC Curve
A graphical plot that illustrates the diagnostic ability of a binary classifier system.
AUC (Area Under the ROC Curve)
A performance measurement for classification problems at various threshold settings.
Loss Function
A method of evaluating how well specific algorithm models the given data.
Gradient Descent
An optimization algorithm used to minimize some function by iteratively moving in the direction of steepest descent.
Backpropagation
A method used in artificial neural networks to calculate a gradient that is needed in the calculation of the weights to be used in the network.
Epoch
One complete pass through the training dataset.
Batch Size
The number of training examples utilized in one iteration.
Dropout
A regularization technique for reducing overfitting in neural networks.
Active Learning
A training approach where the algorithm can choose the data it learns from. It queries the most informative and relevant examples to learn faster with fewer data (frequently with human-in-the-loop). |
Transfer Learning
Involves taking a model pre-trained or trained on a given task and fine-tuning it for a different task, effective in reducing the need for large labeled datasets.
Federated Learning
A technique for training algorithms across decentralized devices or servers holding local data samples, without exchanging them. Enhances privacy and reduces data centralization.
Fine-Tuning
Fine-tuning in AI refers to taking a pre-trained model (like a foundation model) and further training it on a more specific dataset to specialize its abilities, usually through supervised or reinforcement learning.
Explainable AI (XAI)
Methods and techniques in AI that make the results of the solution understandable by humans.
AI Ethics
The branch of ethics that focuses on the moral and ethical implications of AI and related technologies.
AI Governance
The framework that ensures the development and deployment of AI is aligned with organizational values and objectives.
AI Act
Legislation proposed by the European Union aimed at regulating AI to ensure it is trustworthy and respects fundamental rights.
Multimodal AI
AI that can process and analyze multiple types of data inputs (e.g., text, images, and audio).
Self-Supervised Learning
A type of unsupervised learning where the data itself provides the supervision.
Zero-Shot Learning
The ability of a model to recognize objects it has never seen before.
Few-Shot Learning
A type of machine learning problem where the algorithm is trained to learn information about a category from a small number of training examples.
Adversarial Machine Learning
Techniques that attempt to fool models by supplying deceptive input.
Robustness
The ability of an AI system to continue to function properly in the presence of invalid, incomplete, or unexpected inputs.
Scalability
The capability of an AI system to handle growing amounts of work or its potential to be enlarged to accommodate growth.
Edge AI
AI algorithms processed locally on a hardware device rather than in a centralized data center.
Cloud AI
AI services provided through cloud computing.
Cognitive Computing
Technology platforms that simulate human thought processes in a computerized model.
Speech and Audio Processing
Involves the analysis and interpretation of sound, such as speech recognition, voice synthesis, and audio enhancement using AI techniques.
Computer Vision
A field of AI focused on enabling machines to interpret and understand the visual world from digital images and videos, used in object detection, image classification, etc.
Content Generation
The use of AI to automatically create content, including text, images, videos, and music, often leveraging GANs, LLMs, and other generative models.
Robotics
Involves the use of AI to control and coordinate robots, enabling them to perform complex tasks autonomously or semi-autonomously, often used in manufacturing and healthcare.
Complex Systems Analysis
This involves using AI techniques to understand, model, and predict the behaviour of complex systems. AI can process vast datasets and model intricate interactions within these systems.
Machine Translation
The application of AI to automatically translate text or speech from one language to another, utilizing NLP and deep learning techniques.