AI and Machine Learning

Artificial Intelligence (AI) Computer systems that perform tasks typically requiring human intelligence.

Machine Learning (ML) Systems that learn patterns from data rather than being explicitly programmed.

Deep Learning Machine learning using neural networks with many layers.

Large Language Model (LLM) AI trained on vast text to understand and generate language. Includes ChatGPT, Claude.

Generative AI AI that creates new content — text, images, code, video.

Natural Language Processing (NLP) AI that understands and processes human language.

Computer Vision AI that interprets images and video.

Neural Network Computing architecture loosely inspired by brain structure, used in most modern AI.

Model Development

Training Teaching an AI model using data. Expensive and time-consuming.

Inference Using a trained model to make predictions. Fast and cheap per use.

Fine-Tuning Adjusting a pre-trained model with specific data for better performance on a task.

Prompt Engineering Crafting effective instructions for generative AI to produce desired outputs.

Training Data Data used to teach the model. Quality and relevance directly impact results.

Test Data Data used to evaluate model performance. Should be separate from training data.

Overfitting When a model learns training data too specifically and fails on new data.

Model Capabilities and Limitations

Accuracy How often the model is correct. Usually expressed as a percentage.

Precision Of the items the model identified as positive, how many actually were.

Recall Of all actual positives, how many did the model identify.

Hallucination When AI generates plausible-sounding but false information.

Edge Case Unusual situation at the boundaries of what the model handles well.

Bias Systematic unfairness in model outputs, often reflecting biased training data.

Drift Degradation of model performance over time as conditions change.

Data Concepts

Structured Data Data organized in tables with clear rows and columns. Databases, spreadsheets.

Unstructured Data Data without predefined format. Text documents, images, audio.

Data Pipeline Systems that move and transform data from sources to destinations.

Data Labeling Adding tags or categories to data so AI can learn from it.

Data Quality Accuracy, completeness, and reliability of data.

RAG (Retrieval-Augmented Generation) Connecting AI to external data sources so it can retrieve information before responding.

Embeddings Numerical representations of data that capture meaning and relationships.

Implementation

Pilot Controlled test of an AI system before full deployment.

Production Live deployment of AI system in actual business operations.

Proof of Concept (POC) Demonstration that an approach can work. Earlier and simpler than pilot.

MVP (Minimum Viable Product) Simplest version that delivers value. Ship fast, learn, iterate.

Integration Connecting AI systems to existing business systems.

MLOps Practices for deploying and maintaining machine learning in production.

Scalability Ability to handle increased load without degradation.

Monitoring Tracking AI system performance in production.

Organizational

Business Case Justification for investment showing costs, benefits, and risks.

Stakeholder Anyone affected by or interested in an AI initiative.

Executive Sponsor Senior leader accountable for initiative success.

Change Management Activities to help people adopt new ways of working.

Center of Excellence (CoE) Centralized team providing expertise and standards.

Vendors and Procurement

SaaS (Software as a Service) Software delivered over the internet, typically subscription-based.

API (Application Programming Interface) Way for systems to communicate. How you access AI services.

SLA (Service Level Agreement) Contractual guarantee of service quality (uptime, response time).

Vendor Lock-in Difficulty switching away from a vendor due to dependencies.

Total Cost of Ownership (TCO) All costs over time, not just initial purchase.

Governance and Risk

AI Governance Policies, processes, and structures for responsible AI use.

AI Ethics Principles guiding responsible development and use of AI.

Algorithmic Bias When AI systems produce unfair outcomes for certain groups.

Explainability Ability to understand why AI made a particular decision.

Transparency Being open about AI use, capabilities, and limitations.

Audit Trail Record of what AI did and why, enabling review.

Regulation

GDPR (General Data Protection Regulation) European data protection regulation with strict requirements.

EU AI Act European regulation classifying AI systems by risk and imposing requirements.

CCPA (California Consumer Privacy Act) California data privacy law giving consumers rights over personal data.

Right to Explanation Legal requirement (in some contexts) to explain automated decisions.

High-Risk AI AI systems subject to stricter regulation due to potential impact.

Security

Adversarial Attack Inputs designed to fool AI systems.

Prompt Injection Malicious instructions hidden in inputs to manipulate AI.

Data Poisoning Corrupting training data to compromise model behavior.

Model Theft Extracting proprietary models through repeated queries.

Roles

Data Scientist Builds and trains models. Analyzes data for patterns.

ML Engineer (Machine Learning Engineer) Focuses on production deployment of models.

Data Engineer Builds and maintains data pipelines and infrastructure.

MLOps Engineer Specializes in deploying and operating ML systems.

AI Product Manager Connects business needs to AI capabilities. Prioritizes and coordinates.

AI Architect Designs overall AI system architecture.