AI is the field of creating computer systems that can perform tasks associated with human intelligence

Focuses on perceiving, learning, reasoning, deciding, and acting toward goals

Automate intelligent behavior

Augment human capabilities (decision support, prediction, personalization)

Build systems that adapt to changing environments and data

Perception (sensing the world via data)

Representation (encoding knowledge/features)

Learning (improving from data/experience)

Reasoning (drawing conclusions, planning)

Interaction (communicating with humans or other systems)

Autonomy (acting with limited supervision)

Narrow AI (ANI)

General AI (AGI)

Superintelligent AI (ASI)

Reactive Machines

Limited Memory

Theory of Mind (conceptual)

Self-Aware (conceptual)

Learns patterns from data instead of explicit rules

Supervised Learning

Unsupervised Learning

Semi-supervised Learning

Reinforcement Learning (RL)

Uses multi-layer neural networks to learn representations

Excels in vision, speech, language, and complex pattern recognition

Typical models: CNNs (images), RNNs/Transformers (sequences)

Uses explicit rules, logic, and knowledge graphs

Strengths: interpretability, formal reasoning

Limitations: brittle rules, costly knowledge engineering

Combines ML/deep learning with symbolic reasoning

Goal: data-driven learning + structured reasoning + better reliability

Produces new content (text, images, audio, code)

Often based on large foundation models trained on broad datasets

Common uses: drafting, summarization, design prototyping, coding assistance

Define objective, constraints, success metrics

Choose task type (classification, prediction, generation, control)

Collection (logs, sensors, text corpora, images)

Labeling/annotation (when needed)

Data quality concerns (noise, imbalance, missing values)

Feature engineering (traditional ML) or representation learning (deep learning)

Model selection and training

Hyperparameter tuning

Offline metrics (accuracy, precision/recall, F1, AUC, RMSE)

Robustness tests (edge cases, distribution shift)

Fairness and bias checks

Integration into products/workflows (APIs, apps, devices)

Monitoring (performance drift, failures, data changes)

Iteration (retraining, model updates)

Understanding and generating human language

Tasks: translation, summarization, question answering, sentiment analysis

Interpreting images and video

Tasks: detection, segmentation, OCR, medical imaging analysis

Speech recognition, speaker identification, text-to-speech

Perception + planning + control in physical environments

Navigation, manipulation, human-robot interaction

Scheduling, route planning, resource allocation

Personalized ranking and suggestions

Uses behavior signals, content features, and contextual data

Logic, ontologies, knowledge graphs, causal reasoning

Search ranking and autocomplete

Virtual assistants and chatbots

Photo organization and face recognition

Smart home devices (voice control, automation)

Customer support automation and triage

Forecasting (demand, inventory)

Fraud detection and risk scoring

Process automation (document understanding, claims processing)

Imaging diagnostics assistance (X-ray/CT/MRI support)

Clinical decision support and risk prediction

Drug discovery and protein modeling

Patient monitoring and virtual care

Credit scoring and underwriting support

Algorithmic trading and market analysis

Anti-money laundering and anomaly detection

Personalized learning paths and tutoring systems

Automated grading support and feedback

Accessibility tools (captioning, reading assistance)

Predictive maintenance for equipment

Quality inspection with vision systems

Robotics for assembly and warehousing

Route optimization and fleet management

Driver assistance and autonomous driving components

Warehouse automation

Malware detection and behavior analysis

Identity verification and access control

Threat intelligence summarization

Content recommendation (music/video)

Generative tools for art, video, and music

Game AI and procedural generation

Document processing and service routing

Infrastructure monitoring

Public safety analytics (requires careful governance)

Requires sufficient, representative, and high-quality data

Sensitive to dataset shift (changing real-world conditions)

Errors on edge cases; brittle behavior outside training distribution

Overconfidence and calibration problems

Hard to explain decisions for complex models (black-box concerns)

Can amplify societal biases present in data

Unequal performance across groups if not addressed

Risks of data leakage and model inversion

Adversarial attacks and prompt injection (for generative systems)

Compute, energy use, infrastructure complexity

Overreliance and automation bias

Need for training, change management, and clear accountability

Fairness, accountability, transparency, privacy, and safety

Human-in-the-loop for high-stakes decisions

Testing for robustness, misuse, and failure modes

Incident response and audit trails

Data protection laws (e.g., GDPR-like requirements)

Sector regulations (health, finance, employment)

Model documentation (datasheets, model cards)

AI is the broader goal; ML is a major method to achieve it

Deep learning is a subset of ML using deep neural networks

Automation can be rule-based without learning; AI often involves learning/adaptation

Data science focuses on extracting insights; AI focuses on building intelligent behavior in systems