Artificial Intelligence: Part 4

Binary Classification

Binary classification is a task where an AI system chooses between two categories. Examples include spam or not spam, fraud or not fraud, disease or no disease, approve or reject. It sounds simple, but the consequences can be serious. In high-stakes settings, we need to understand not only the answer but also the confidence, error rate and impact of mistakes.

Black Box

A black-box AI system is one whose internal reasoning is difficult for humans to understand. Many modern AI models are powerful but not easily explainable. This creates a problem in areas such as healthcare, finance and law, where people may need to know why a decision was made. Black-box systems can be useful, but they require careful testing and oversight.

BLEU Score

BLEU is a score traditionally used to evaluate machine translation. It compares an AI-generated translation with one or more human translations. A higher BLEU score usually means the machine translation is closer to the reference text. However, like many metrics, it is imperfect. A translation can score well but still sound unnatural, or score poorly while being acceptable to a human reader.

Bot

A bot is a software program that performs automated tasks. Some bots are simple, such as those that send reminders or answer basic customer-service questions. AI-powered bots are more flexible and can understand natural language. Bots can be helpful, but they can also be used for spam, scams, misinformation or fake social-media activity.

Bounding Box

A bounding box is a rectangle drawn around an object in an image so an AI system can learn to detect it. For example, in a traffic image, boxes may be drawn around cars, pedestrians, bicycles and signs. Bounding boxes are commonly used in computer vision. They show the AI not just what is in an image, but where it is.

BPE — Byte Pair Encoding

Byte Pair Encoding, or BPE, is a method used to break text into smaller units called tokens. These tokens may be whole words, parts of words or characters. Language models do not usually read text exactly as humans do; they process tokens. Tokenisation affects cost, speed and how well a model handles unusual words, names or languages.

Boosting

Boosting is a machine-learning technique that combines many weaker models to create a stronger one. Each new model focuses more on the mistakes made by earlier models. Over time, the combined system becomes more accurate. Boosting is widely used in structured-data problems, such as risk scoring, forecasting and classification.

Business Intelligence

Business intelligence uses data to help organisations understand performance and make decisions. Traditional business intelligence often involves dashboards and reports. AI can enhance business intelligence by finding patterns, producing forecasts, summarising trends and suggesting actions. However, AI-generated business insights still need human judgement and domain knowledge.