AI & Machine Learning Basics

AI & Machine Learning Basics

What is Artificial Intelligence?

Artificial Intelligence (AI) is the simulation of human intelligence in machines — enabling them to learn, reason, perceive, and interact. AI systems range from narrow task-specific tools to increasingly general-purpose reasoning systems.

Three levels of AI:

• Narrow AI (ANI) — Designed for specific tasks (ChatGPT, image classifiers, recommendation engines)
• General AI (AGI) — Hypothetical: performs any intellectual task a human can
• Super AI (ASI) — Hypothetical: surpasses human intelligence across all domains

What is Machine Learning?

Machine Learning (ML) is a subset of AI where systems learn patterns from data rather than following explicit rules. The model improves with experience.

Types of Machine Learning

Supervised Learning — Learns from labeled data (input → output pairs)

• Examples: spam detection, image classification, price prediction
• Algorithms: Linear/Logistic Regression, Decision Trees, SVM, Neural Networks

Unsupervised Learning — Finds patterns in unlabeled data

• Examples: customer segmentation, anomaly detection, topic modeling
• Algorithms: K-Means, DBSCAN, PCA, Autoencoders

Reinforcement Learning — Agent learns by interacting with an environment, receiving rewards/penalties

• Examples: game-playing AI (AlphaGo), robotics, trading bots
• Key concepts: Agent, Environment, State, Action, Reward, Policy

Self-Supervised Learning — Model generates its own labels from input structure

• Examples: GPT (predicts next token), BERT (predicts masked tokens), CLIP
• Foundation of modern LLMs

Deep Learning

Deep Learning uses multi-layer neural networks to learn hierarchical representations:

Architecture	Use Case
CNN (Convolutional)	Images, video
RNN / LSTM / GRU	Sequences, time series
Transformer	Text, code, multimodal
Diffusion Models	Image/audio generation
GAN	Synthetic data generation

The Transformer Revolution

Transformers (introduced in “Attention Is All You Need”, 2017) are the backbone of modern AI:

• Self-attention mechanism allows the model to weigh relationships between all tokens simultaneously
• Scales efficiently with data and compute
• Powers GPT, BERT, T5, LLaMA, Gemini, Claude, and virtually all modern LLMs

Large Language Models (LLMs)

LLMs are transformer-based models trained on massive text corpora. Key concepts:

• Pre-training — Learn language patterns from billions of tokens
• Fine-tuning — Adapt to specific tasks with smaller labeled datasets
• RLHF — Reinforcement Learning from Human Feedback — aligns model to human preferences
• Prompt engineering — Crafting inputs to guide model behavior
• RAG — Retrieval-Augmented Generation — augment LLM with external knowledge at inference time
• Context window — Maximum tokens the model can process at once (4K → 1M+ tokens in 2026)

AI in 2026: Key Trends

Agentic AI — Models that plan, use tools, and execute multi-step tasks autonomously (see AI Agents)

Multimodal Models — Process text, images, audio, video, and code together (GPT-4o, Gemini 2.0, Claude 3.5)

Reasoning Models — Dedicated “thinking” before answering (OpenAI o1/o3, DeepSeek R1)

Small Language Models (SLMs) — Efficient models for edge/on-device use (Phi-4, Gemma 3, Llama 3.2 1B)

AI Coding Assistants — GitHub Copilot, Cursor, Kiro — deeply integrated into developer workflows

Open Source AI — LLaMA 4, Mistral, Qwen 3, DeepSeek V3 — competitive with closed models

AI vs ML vs DL

	AI	ML	DL
Scope	Broadest	Subset of AI	Subset of ML
Approach	Rules + learning	Statistical learning	Neural networks
Data needs	Varies	Moderate	Large
Interpretability	Varies	Often interpretable	Often black-box
Examples	Expert systems, LLMs	Random forests, SVM	GPT, ResNet, DALL-E

Core ML Workflow

1. Define problem & success metrics
2. Collect & label data
3. Exploratory Data Analysis (EDA)
4. Feature engineering & preprocessing
5. Model selection & training
6. Evaluation (accuracy, F1, AUC, etc.)
7. Hyperparameter tuning
8. Deployment & monitoring
9. Iterate

Key Tools & Frameworks

Category	Tools
Languages	Python, R, Julia
ML Libraries	scikit-learn, XGBoost, LightGBM
Deep Learning	PyTorch, TensorFlow/Keras, JAX
LLM Frameworks	Hugging Face Transformers, LangChain, LlamaIndex
Data	Pandas, NumPy, Polars, DuckDB
Visualization	Matplotlib, Seaborn, Plotly
Experiment Tracking	MLflow, Weights & Biases, Neptune
Deployment	FastAPI, BentoML, Triton, vLLM
Cloud AI	AWS SageMaker, Azure ML, Google Vertex AI

Real-World Applications

• Healthcare — Medical imaging, drug discovery, clinical decision support
• Finance — Fraud detection, algorithmic trading, credit scoring
• Retail — Recommendation systems, demand forecasting, dynamic pricing
• Automotive — Autonomous driving, predictive maintenance
• Software — Code generation, bug detection, automated testing
• Science — Protein folding (AlphaFold), climate modeling, materials discovery

Evaluation Metrics

Task	Metrics
Classification	Accuracy, Precision, Recall, F1, AUC-ROC
Regression	MAE, MSE, RMSE, R²
Generation	BLEU, ROUGE, BERTScore, human eval
Ranking	NDCG, MRR
Clustering	Silhouette score, Davies-Bouldin

Further Reading

• AI Agents Overview
• Agentic AI Concepts
• Spectrum of Agency
• Top AI Tools 2026
• AI Models 2026