# The Executive's AI Lexicon

> 100 terms every executive needs to lead an AI program with confidence — from LLMs, tokens, and context windows to RAG, agentic AI, and neural rendering. Searchable, A–Z filterable, each with an animated schematic and the strategic impact.

**Canonical URL:** https://www.andekian.com/ai-lexicon  
**Author / Site:** Stephen Andekian — https://www.andekian.com

## All 100 Terms

1. [LLM — Large Language Model](https://www.andekian.com/ai-lexicon/llm): The computational foundation behind all generative AI systems.
2. [Hallucination — Confidence Without Accuracy](https://www.andekian.com/ai-lexicon/hallucination): Confident but factually incorrect outputs generated by a model.
3. [Token — Unit Of AI Processing](https://www.andekian.com/ai-lexicon/token): The fundamental unit of data processing in AI — roughly 0.75 words.
4. [Context Window — Operational Memory Limit](https://www.andekian.com/ai-lexicon/context-window): The operational memory limit of an AI during an active session.
5. [Fine-Tuning — Domain-Specific Mastery](https://www.andekian.com/ai-lexicon/fine-tuning): Adapting a pre-trained model for domain-specific mastery using internal data.
6. [RLHF — Reinforcement Learning From Human Feedback](https://www.andekian.com/ai-lexicon/rlhf): Aligns raw model capabilities with human values, safety criteria, and intent.
7. [SLMs & Distillation — Compression · Speed · Deployment](https://www.andekian.com/ai-lexicon/slms-and-distillation): Transferring capabilities from massive frontier models into compact architectures.
8. [RAG — Retrieval-Augmented Generation](https://www.andekian.com/ai-lexicon/rag): Dynamically connecting live internal databases directly to an LLM's prompt window.
9. [Chain of Thought — Sequential Reasoning Engine](https://www.andekian.com/ai-lexicon/chain-of-thought): Prompting or forcing a model to execute complex reasoning before responding.
10. [Weights & Parameters — Learned Intelligence As Math](https://www.andekian.com/ai-lexicon/weights-and-parameters): The mathematical values within a neural network that dictate connection strength.
11. [Validation Loss — Training Health Indicator](https://www.andekian.com/ai-lexicon/validation-loss): Measures how well an AI generalizes to data it hasn't seen during training.
12. [Agentic AI — Autonomous Workflow Execution](https://www.andekian.com/ai-lexicon/agentic-ai): Systems engineered to act autonomously, plan complex tasks, and use external tools.
13. [Inference — Runtime AI Execution](https://www.andekian.com/ai-lexicon/inference): The process of running a trained model to generate outputs from new inputs.
14. [Multimodal AI — Text-Image-Audio Reasoning](https://www.andekian.com/ai-lexicon/multimodal-ai): Models that process and reason across text, images, audio, and video simultaneously.
15. [Embeddings — Meaning Encoded As Vectors](https://www.andekian.com/ai-lexicon/embeddings): Numerical vector representations that encode the semantic meaning of text or images.
16. [Vector Database — Stores Vector Embeddings](https://www.andekian.com/ai-lexicon/vector-database): Specialized infrastructure optimized for storing and querying high-dimensional embedding vectors.
17. [Transformer Architecture — Modern LLM Foundation](https://www.andekian.com/ai-lexicon/transformer-architecture): The foundational neural network design behind virtually all modern large language models.
18. [Attention Mechanism — Prioritizes Relevant Context](https://www.andekian.com/ai-lexicon/attention-mechanism): Allows models to dynamically focus on the most relevant tokens for any given output.
19. [Positional Encoding — Sequence Awareness System](https://www.andekian.com/ai-lexicon/positional-encoding): Injects sequence order information into transformer models that process tokens in parallel.
20. [Pretraining — Large-Scale Model Learning](https://www.andekian.com/ai-lexicon/pretraining): Large-scale unsupervised learning on massive datasets to build general capabilities.
21. [Supervised Learning — Labeled Training Data](https://www.andekian.com/ai-lexicon/supervised-learning): Training a model using labeled input-output pairs provided by humans.
22. [Unsupervised Learning — Pattern Discovery Process](https://www.andekian.com/ai-lexicon/unsupervised-learning): Discovering patterns and structure in data without predefined labels or targets.
23. [Self-Supervised Learning — Model Creates Labels](https://www.andekian.com/ai-lexicon/self-supervised-learning): The model generates its own training signal by predicting masked or hidden parts of data.
24. [Transfer Learning — Reuses Learned Intelligence](https://www.andekian.com/ai-lexicon/transfer-learning): Applying knowledge gained from one domain to accelerate learning in another.
25. [Few-Shot Learning — Minimal Example Training](https://www.andekian.com/ai-lexicon/few-shot-learning): Model generalizes to new tasks from only a handful of labeled examples.
26. [Zero-Shot Learning — No Training Examples](https://www.andekian.com/ai-lexicon/zero-shot-learning): Model performs tasks it was never explicitly trained on, using generalized reasoning.
27. [One-Shot Learning — Single-Example Learning](https://www.andekian.com/ai-lexicon/one-shot-learning): Recognizes patterns and generalizes from a single training example.
28. [Prompt Engineering — Instruction Optimization](https://www.andekian.com/ai-lexicon/prompt-engineering): The discipline of crafting inputs that reliably elicit desired model behaviors and outputs.
29. [Prompt Tuning — Prompt-Level Optimization](https://www.andekian.com/ai-lexicon/prompt-tuning): Optimizing a small set of learnable prompt tokens prepended to frozen model inputs.
30. [Instruction Tuning — Human-Guided Refinement](https://www.andekian.com/ai-lexicon/instruction-tuning): Training models on paired instruction-response datasets to improve instruction following.
31. [Alignment — Human-Value Matching](https://www.andekian.com/ai-lexicon/alignment): Engineering AI systems whose goals, behaviors, and values match human intent.
32. [AI Safety — Risk Mitigation Systems](https://www.andekian.com/ai-lexicon/ai-safety): A multi-disciplinary field focused on preventing unintended harms from AI systems.
33. [Emergent Behavior — Unexpected Model Abilities](https://www.andekian.com/ai-lexicon/emergent-behavior): Capabilities that appear unexpectedly in larger models not present in smaller ones.
34. [Scaling Laws — Bigger Models Improve](https://www.andekian.com/ai-lexicon/scaling-laws): Empirical relationships predicting how model performance improves with compute, data, and parameters.
35. [Frontier Model — State-Of-The-Art AI](https://www.andekian.com/ai-lexicon/frontier-model): The highest-capability AI model available at a given point in time.
36. [Open Weights — Public Model Parameters](https://www.andekian.com/ai-lexicon/open-weights): Model parameters made publicly available for inspection, download, and modification.
37. [Closed Weights — Restricted Parameters](https://www.andekian.com/ai-lexicon/closed-weights): Proprietary model parameters accessible only through a vendor API.
38. [Quantization — Reduced Precision Models](https://www.andekian.com/ai-lexicon/quantization): Reducing model weight precision from 32-bit floats to smaller formats like INT8 or INT4.
39. [Model Pruning — Removes Unnecessary Weights](https://www.andekian.com/ai-lexicon/model-pruning): Removing low-importance weights or neurons from a trained model to reduce size.
40. [Sparse Models — Partial Network Activation](https://www.andekian.com/ai-lexicon/sparse-models): Architectures where only a small fraction of neurons activate for any given input.
41. [Mixture of Experts — Specialized Sub-Model Routing](https://www.andekian.com/ai-lexicon/mixture-of-experts): Architecture routing each input to a small subset of specialized sub-networks.
42. [Synthetic Data — AI-Generated Datasets](https://www.andekian.com/ai-lexicon/synthetic-data): AI-generated training data used to augment or replace real-world datasets.
43. [Dataset Curation — Refined Training Inputs](https://www.andekian.com/ai-lexicon/dataset-curation): Systematic filtering, deduplication, and quality scoring of training datasets.
44. [Benchmarking — Standardized AI Evaluation](https://www.andekian.com/ai-lexicon/benchmarking): Standardized evaluation frameworks measuring AI performance on defined tasks.
45. [Overfitting — Poor Generalization](https://www.andekian.com/ai-lexicon/overfitting): Model memorizes training data patterns rather than learning generalizable rules.
46. [Underfitting — Insufficient Learning](https://www.andekian.com/ai-lexicon/underfitting): Model is too simple to capture the underlying patterns in training data.
47. [Gradient Descent — Optimization Algorithm](https://www.andekian.com/ai-lexicon/gradient-descent): Iterative optimization algorithm that updates model weights to minimize prediction error.
48. [Backpropagation — Neural Weight Adjustment](https://www.andekian.com/ai-lexicon/backpropagation): Algorithm computing gradients of the loss function through a neural network in reverse.
49. [Epoch — Complete Training Cycle](https://www.andekian.com/ai-lexicon/epoch): One complete pass through the entire training dataset during model optimization.
50. [Hyperparameters — Training Configuration Settings](https://www.andekian.com/ai-lexicon/hyperparameters): Configuration settings controlling the training process rather than learned from data.
51. [Loss Function — Measures Prediction Error](https://www.andekian.com/ai-lexicon/loss-function): Mathematical function measuring the difference between model predictions and true labels.
52. [Neural Network — Layered AI Architecture](https://www.andekian.com/ai-lexicon/neural-network): Computational graph of interconnected artificial neurons organized in layers.
53. [Deep Learning — Multi-Layer Neural Training](https://www.andekian.com/ai-lexicon/deep-learning): Machine learning using neural networks with many successive layers of representation.
54. [Diffusion Model — Generative Image Architecture](https://www.andekian.com/ai-lexicon/diffusion-model): Generative model learning to reverse a noise-addition process to create new content.
55. [Foundation Model — Large Generalized Model](https://www.andekian.com/ai-lexicon/foundation-model): Large general-purpose model pretrained at scale, adaptable to many downstream tasks.
56. [Semantic Search — Meaning-Based Retrieval](https://www.andekian.com/ai-lexicon/semantic-search): Retrieval systems that match queries based on meaning rather than keyword overlap.
57. [Hybrid Search — Vector + Keyword Search](https://www.andekian.com/ai-lexicon/hybrid-search): Combining semantic vector search with traditional keyword-based retrieval.
58. [Knowledge Graph — Connected Entity Networks](https://www.andekian.com/ai-lexicon/knowledge-graph): Structured network representing entities and their typed relationships.
59. [Chunking — Document Segmentation Process](https://www.andekian.com/ai-lexicon/chunking): Dividing documents into smaller segments for efficient embedding and retrieval.
60. [Grounding — Source-Connected Outputs](https://www.andekian.com/ai-lexicon/grounding): Connecting AI outputs to verified, retrievable source information.
61. [Context Injection — Dynamic Information Insertion](https://www.andekian.com/ai-lexicon/context-injection): Dynamically inserting retrieved information into a model's prompt at inference time.
62. [Similarity Search — Finds Related Meaning](https://www.andekian.com/ai-lexicon/similarity-search): Finding the most semantically similar items to a query in a vector space.
63. [Vector Search — Embedding-Based Retrieval](https://www.andekian.com/ai-lexicon/vector-search): Querying a vector database to retrieve embeddings closest to a query vector.
64. [Knowledge Cutoff — Training Data Endpoint](https://www.andekian.com/ai-lexicon/knowledge-cutoff): The date beyond which a model has no training data and cannot know recent events.
65. [Context Compression — Smaller Context Footprint](https://www.andekian.com/ai-lexicon/context-compression): Reducing token count of long contexts while preserving semantic content.
66. [Citation Grounding — Traceable Source Linking](https://www.andekian.com/ai-lexicon/citation-grounding): Linking specific AI claims to verifiable source documents with traceable references.
67. [Memory Persistence — Retained AI State](https://www.andekian.com/ai-lexicon/memory-persistence): Storing and retrieving information across multiple AI sessions or user interactions.
68. [Long-Term Memory — Persistent Contextual Storage](https://www.andekian.com/ai-lexicon/long-term-memory): Persistent storage of information beyond a single context window or session.
69. [Short-Term Memory — Active Session Awareness](https://www.andekian.com/ai-lexicon/short-term-memory): The active information available within the current context window.
70. [Retrieval Pipeline — Information Retrieval Flow](https://www.andekian.com/ai-lexicon/retrieval-pipeline): The full sequence of steps from user query to returned context in a RAG system.
71. [Retrieval Precision — Accurate Information Fetching](https://www.andekian.com/ai-lexicon/retrieval-precision): The fraction of retrieved results that are actually relevant to the query.
72. [Retrieval Recall — Broad Knowledge Retrieval](https://www.andekian.com/ai-lexicon/retrieval-recall): The fraction of all relevant documents that are successfully retrieved.
73. [Hallucination Mitigation — Reduces False Outputs](https://www.andekian.com/ai-lexicon/hallucination-mitigation): Techniques and architectures designed to reduce AI-generated factual errors.
74. [AI Agent — Autonomous AI Operator](https://www.andekian.com/ai-lexicon/ai-agent): An AI system capable of autonomously perceiving its environment and taking goal-directed actions.
75. [Multi-Agent System — Collaborative AI Agents](https://www.andekian.com/ai-lexicon/multi-agent-system): Networks of AI agents collaborating to solve tasks exceeding any single agent's capacity.
76. [Tool Calling — External Tool Usage](https://www.andekian.com/ai-lexicon/tool-calling): Enabling AI models to invoke external tools like search, calculators, or APIs.
77. [Function Calling — Structured API Execution](https://www.andekian.com/ai-lexicon/function-calling): Structured mechanism for LLMs to trigger specific code functions with typed parameters.
78. [Autonomous Planning — Independent Task Sequencing](https://www.andekian.com/ai-lexicon/autonomous-planning): AI systems independently generating and sequencing action plans to achieve goals.
79. [Reflection Loop — Self-Review Mechanism](https://www.andekian.com/ai-lexicon/reflection-loop): Agent reviewing and critiquing its own outputs before finalizing results.
80. [Recursive Reasoning — Multi-Pass Problem Solving](https://www.andekian.com/ai-lexicon/recursive-reasoning): Applying the same reasoning process repeatedly at increasing levels of abstraction.
81. [Self-Correction — Autonomous Error Fixing](https://www.andekian.com/ai-lexicon/self-correction): AI system detecting and autonomously fixing errors in its own generated outputs.
82. [Chain-of-Verification — Step-By-Step Validation](https://www.andekian.com/ai-lexicon/chain-of-verification): Generating a response then systematically verifying each factual claim independently.
83. [Tree of Thoughts — Branching Reasoning Framework](https://www.andekian.com/ai-lexicon/tree-of-thoughts): Reasoning strategy exploring multiple solution branches and selecting the most promising.
84. [ReAct Framework — Reasoning Plus Acting](https://www.andekian.com/ai-lexicon/react-framework): Interleaved reasoning and acting — model thinks, acts, observes, then thinks again.
85. [Planner Model — Task Sequencing Intelligence](https://www.andekian.com/ai-lexicon/planner-model): AI system dedicated to breaking down high-level goals into executable sub-task sequences.
86. [Copilot — Human-Assistive AI](https://www.andekian.com/ai-lexicon/copilot): AI system designed to augment human performance rather than replace human decision-making.
87. [Autonomous Execution — Reduced Human Intervention](https://www.andekian.com/ai-lexicon/autonomous-execution): AI completing entire workflows independently without human intervention or approval.
88. [AI Governance — AI Oversight Systems](https://www.andekian.com/ai-lexicon/ai-governance): Policies, processes, and controls ensuring responsible AI development and deployment.
89. [Guardrails — Behavioral Constraints](https://www.andekian.com/ai-lexicon/guardrails): Technical and policy controls constraining AI outputs to acceptable behavior ranges.
90. [Red Teaming — Adversarial AI Testing](https://www.andekian.com/ai-lexicon/red-teaming): Adversarial testing by dedicated teams attempting to break or manipulate AI systems.
91. [Explainable AI (XAI) — Transparent AI Reasoning](https://www.andekian.com/ai-lexicon/explainable-ai-xai): Methods enabling humans to understand why an AI system produced a specific output.
92. [Observability — Production AI Monitoring](https://www.andekian.com/ai-lexicon/observability): Comprehensive monitoring of AI system behavior, performance, and drift in production.
93. [Model Drift — Performance Degradation Over Time](https://www.andekian.com/ai-lexicon/model-drift): Gradual degradation of model performance as real-world conditions diverge from training.
94. [Data Drift — Shifting Input Distributions](https://www.andekian.com/ai-lexicon/data-drift): Shift in the statistical properties of input data over time, affecting model reliability.
95. [Reinforcement Learning — Reward-Based Training](https://www.andekian.com/ai-lexicon/reinforcement-learning): Training paradigm where agents learn through reward signals from environmental feedback.
96. [Constitutional AI — Rule-Based Alignment](https://www.andekian.com/ai-lexicon/constitutional-ai): Alignment technique where models self-critique outputs against an explicit rule set.
97. [Active Learning — Human-Guided Data Labeling](https://www.andekian.com/ai-lexicon/active-learning): Selectively querying human labels for the most informative uncertain training examples.
98. [Latent Space — Hidden Representation Space](https://www.andekian.com/ai-lexicon/latent-space): The compressed, high-dimensional internal representation space of a trained neural network.
99. [AI Inference Engine — Model Execution Infrastructure](https://www.andekian.com/ai-lexicon/ai-inference-engine): Optimized software and hardware infrastructure for serving trained models at production scale.
100. [Neural Rendering — AI-Generated Visual Synthesis](https://www.andekian.com/ai-lexicon/neural-rendering): Using neural networks to generate photorealistic images from learned scene representations.

## Contact

Book a conversation or send an inquiry: https://www.andekian.com/#contact
LinkedIn: https://www.linkedin.com/in/andekian/