Unknown Date

Teminology

A

  • Adaptive Knowledge – AI's ability to modify its responses based on newly acquired data.

  • AI-Powered Assistants – Virtual assistants like ChatGPT, Siri, Alexa, which assist users dynamically.

  • Active Learning – AI technique where models learn continuously from user input or feedback.

  • Agent-Based AI – AI that can plan, execute, and adjust tasks autonomously.

  • API (Application Programming Interface) – A set of protocols allowing software components to communicate.

C

  • Contextual Awareness – AI’s ability to retain and apply past interactions to improve responses.

  • Continual Learning – A model’s ability to adapt and improve over time instead of being static.

  • Conversation Buffer Memory – A memory module that stores user interactions to provide continuity in conversations.

  • Cross-Source Validation – AI verifying information by comparing multiple sources for accuracy.

  • Chain-of-Thought (CoT) – AI reasoning technique where responses are generated step by step.

D

  • Dynamic Self-Improvement – AI’s capability to adjust its responses based on learning from interactions.

  • Document Summarization – AI technique where long documents are condensed into key points.

E

  • Embedding Models – AI representations of text, images, or data in numerical vector format for retrieval.

  • Extractive Summarization – AI method that selects key sentences or phrases from a document to create a summary.

F

  • FAISS (Facebook AI Similarity Search) – A vector database used to store and retrieve similar embeddings efficiently.

  • Feedback Loop – A mechanism where AI learns from user corrections or preferences to refine future outputs.

  • Fine-Tuning – The process of training an AI model on domain-specific data to enhance its performance.

G

  • Graph Database – A database designed to store and manage nodes, edges, and relationships (e.g., Neo4j).

  • Graph Embeddings – Vector representations of graph-based knowledge for AI learning.

H

  • Human-in-the-Loop (HITL) – A system where AI receives human feedback for continuous improvement.

  • Hybrid Retrieval – AI combining semantic, keyword, and structured search techniques for information retrieval.

I

  • Information Retrieval (IR) – AI’s ability to fetch relevant documents and data from external sources.

  • Inference Engine – AI component responsible for making logical decisions based on data.

K

  • Knowledge Expansion – The process where AI adds new verified information to its response generation pipeline.

  • Knowledge Graph – A structured database that stores relationships between entities for AI reasoning.

L

  • LAG (Learning-Augmented Generation) – AI technique that continuously updates itself based on feedback and new knowledge.

  • LLM (Large Language Model) – Advanced AI models like GPT-4, Claude, Llama that generate human-like text.

M

  • Memory-Based Retrieval – AI retrieving past user interactions to personalize and improve responses.

  • Model Drift – A situation where an AI model’s performance degrades over time due to outdated knowledge.

  • Multimodal AI – AI that processes and integrates text, images, audio, and video inputs.

N

  • Neural-Symbolic AI – AI combining machine learning (neural networks) with rule-based logic (symbolic AI).

  • Next-Gen RAG (RAG-2) – Enhanced Retrieval-Augmented Generation (RAG) with multi-step retrieval and summarization.

P

  • Personalized Embeddings – AI-generated representations that adapt to individual user preferences.

  • Prompt Engineering – The practice of designing structured inputs to get optimal AI responses.

Q

  • Query Expansion – AI technique where user queries are reformulated to improve retrieval accuracy.

  • Query Reformulation – Adjusting user input to better match stored knowledge in retrieval-based models.

R

  • RAG (Retrieval-Augmented Generation) – A method where AI retrieves external knowledge before generating responses.

  • RLHF (Reinforcement Learning from Human Feedback) – AI training method where models learn from user feedback.

  • Response Refinement – AI’s ability to improve outputs dynamically based on feedback and retrieval.

S

  • Self-Correction Loops – AI mechanisms that automatically refine responses based on continuous feedback.

  • Semantic Search – AI technique that understands intent rather than relying solely on keyword matches.

  • Structured Learning – AI method that incorporates organized datasets and defined rules into its training process.

  • Symbolic AI – AI approach based on logic, rules, and predefined knowledge representations.

T

  • Task Planning AI – AI that breaks down complex objectives into smaller, manageable steps.

  • Tool-Augmented Generation (TAG) – AI method where models call external APIs, functions, or tools to improve responses.

  • Token Limit – The maximum number of words or characters an AI model can process in a single interaction.

  • Transformer-Based AI – A type of deep learning model used in LLMs for processing and generating text.

V

  • Vector Database – A storage method used to store embeddings and perform similarity searches (e.g., FAISS, Pinecone).

  • Vector Search – AI method of finding similar documents using vectorized representations.

W

  • Weighted Retrieval Ranking – AI ranking system that prioritizes important information sources based on credibility.

  • Workflow Automation AI – AI capable of executing multi-step business or operational tasks autonomously.

← Back to Library