📖 AI Glossary

It allows repetitive tasks to be handled without human hands. With the combination of AI, the scope of tasks that require judgment is expanding.

Autonomous driving is a technology that allows a car to drive itself without human intervention. Waymo's robotaxi has entered commercial operation.

Backtesting evaluates an investment strategy by applying it to historical market data.

Benchmark is a standard test to compare AI performance. This is the scorecard that comes out when a new model is announced, and there are frequent controversies about training for exams.

Chain of thought is a technique that allows AI to answer by writing down the solution process step by step rather than just making a conclusion. Significantly improves accuracy for complex problems.

Chatbot is the general name for a program that responds in a conversational format. Since the emergence of generative AI, it has become a term used to refer to ChatGPT-type conversational AI.

Cloud is a method of renting a server over the Internet rather than purchasing it directly. Most AI learning and services run in the cloud.

Compact language model is a language model whose size has been significantly reduced compared to large models. It is cheap and fast, so it is in the spotlight for specific tasks.

Computer vision is an AI field that understands images and videos. It corresponds to the eyes of face recognition, defect detection, and autonomous driving.

A context window is the amount of content an AI model can remember and refer to within a single conversation.

This program automatically collects data from the web. This is a key issue in copyright disputes surrounding AI learning data collection.

A development platform for NVIDIA GPUs. It is considered a key weapon in NVIDIA's dominance in the AI developer ecosystem.

A cutoff is the point where an AI model’s training data ends, so it may not know later events, prices, or news.

Data Center is a facility with tens of thousands of servers and is the physical basis for AI learning and services. Power and cooling issues have become hot topics in the AI era.

Data labeling is the task of attaching a correct answer mark to the learning data. It's also a large, manual industry behind AI.

Dataset is a bundle of data used for AI learning. Quality and scale determine AI performance and are at the center of copyright debates.

Deep learning is a machine learning technique that uses layered artificial neural networks inspired by the human brain.

Deepfake is fake content that synthesizes the faces and voices of real people using AI. It has become a tool for fraud and disinformation and is subject to regulation.

This method runs AI directly on field devices (cameras, sensors, factory equipment) rather than on a server.

Embedding is a technology that converts the meaning of text into a numeric vector. Sentences with similar meanings have close numbers, which becomes the basis for AI search.