AI Glossary
A guide to some common AI terms you may hear in the media, and elsewhere. If you're working with Blume Consult, you don't need to understand this - because we make the process super simple for you.
Activation Function
A function applied to a neuron's output that helps the network learn complex patterns (e.g., ReLU, sigmoid, tanh).
Agentic AI
A concept in AI where systems can autonomously take actions, adapt to changes, and pursue goals with minimal human intervention.
Algorithm
A set of rules or steps for solving a problem or performing a task in a finite number of steps.
Artificial Intelligence (AI)
A field of computer science focused on making machines capable of performing tasks that typically require human intelligence.
Autoencoder
A neural network designed to learn efficient representations of data by reconstructing its input.
Bias (AI Bias)
Systematic errors in AI outputs caused by biased data or assumptions, often leading to unfair treatment or results.
Classification
A supervised learning task that predicts a discrete label or category, such as spam vs. not spam.
Cloud Computing
Using remote servers on the internet to store, manage, and process data, often for running AI services at scale.
Computer Vision
A field of AI that focuses on enabling computers to interpret and understand visual information from the world.
Confusion Matrix
A table summarizing prediction results, showing counts of true positives, false positives, false negatives, and true negatives.
Convolutional Neural Network (CNN)
A type of neural network especially good at analyzing visual data (images, videos) using convolution layers.
Data Mining
The process of discovering patterns and insights from large datasets using statistical and AI techniques.
Data Preprocessing
Cleaning and transforming raw data into a usable format before feeding it into a model.
Deep Learning (DL)
A subset of machine learning that uses multi-layered neural networks to automatically learn complex patterns from large amounts of data.
Dropout
A regularization method where certain neurons are randomly 'dropped' during training to reduce overfitting.
Edge Computing
Processing data near the source (device or sensor) instead of in a centralized cloud, reducing latency and bandwidth usage.
Embedding
A numerical representation of data (often text) in a lower-dimensional space, capturing semantic relationships.
Epoch
One complete pass through the entire training dataset during the training process.
Ethics in AI
Considerations related to the moral and societal impacts of AI, such as privacy, fairness, and accountability.
Explainable AI (XAI)
AI systems designed to provide human-understandable justifications for their outputs or decisions.
F1 Score
A single metric that balances Precision and Recall, calculated as the harmonic mean of the two.
Feature Extraction
The process of identifying the most important attributes (features) that represent the data effectively.
Generative Adversarial Network (GAN)
Two neural networks (generator and discriminator) competing against each other to produce realistic synthetic data.
Gradient Descent
An optimization algorithm that iteratively adjusts model parameters to minimize the loss function.
Hyperparameters
Settings in a model (like learning rate or number of layers) that must be specified before training and can significantly affect performance.
Inference
Using a trained model to make predictions on new, unseen data.
Learning Rate
A hyperparameter that controls how much model parameters are adjusted during training with each step.
LLM (Large Language Model)
A type of AI model trained on massive amounts of text to understand and generate human language.
Loss Function (Cost Function)
A measure of how far the model's predictions are from the target values. Lower loss indicates better performance.
Machine Learning (ML)
A subset of AI that enables machines to learn from data and improve their performance over time without being explicitly programmed.
Model
A mathematical representation of a real-world process, learned from data and used to make predictions or decisions.
Natural Language Processing (NLP)
A branch of AI that enables machines to understand, interpret, and generate human language.
Neural Network
A computing system inspired by the structure of the brain. Consists of interconnected nodes (neurons) that process information.
Neuron (Node)
A fundamental unit in a neural network. It takes inputs, applies weights and biases, and uses an activation function to produce an output.
Overfitting
When a model learns the training data too well, including noise, causing it to perform poorly on new, unseen data.
Pre-trained Model
A model that’s already been trained on a large dataset and can be fine-tuned for a specific task.
Precision
The proportion of correct positive predictions out of all positive predictions. Indicates accuracy among identified positives.
Prompt Engineering
Crafting or refining prompts to guide large language models (LLMs) toward more accurate or relevant outputs.
Recurrent Neural Network (RNN)
A type of neural network designed for sequential data, where connections form loops allowing information to persist over time.
Regression
A supervised learning task that predicts a continuous value, such as prices or temperatures.
Reinforcement Learning
A machine learning technique where an agent learns to make decisions by performing actions and receiving rewards or penalties.
ROC Curve (Receiver Operating Characteristic)
A plot that shows the performance of a classification model at all classification thresholds.
Supervised Learning
A machine learning approach that uses labeled examples to train models, helping them learn to predict labels for new data.
Tokenization
The process of splitting text into smaller units (tokens) such as words or subwords for NLP tasks.
Training
The process of feeding data into a model so it can learn patterns and relationships.
Transformer
A neural network architecture particularly effective for NLP tasks, relying on attention mechanisms to process sequences in parallel.
Underfitting
When a model is too simple and fails to capture the underlying trends in the data, leading to poor performance on both training and test sets.
Unsupervised Learning
A machine learning approach that deals with unlabeled data, discovering hidden patterns or groupings without explicit instructions.
Zero-Shot Learning
A technique where a model can recognize things it has never seen before by leveraging semantic information.