Skip to content
FachwerxLangdock Academy
Back to chapters

What is an LLM?

From traditional AI to language models

Traditional AI vs. LLMs

How LLMs differ from earlier rule-based and narrow ML systems.

Generative AI

LLMs that understand and generate natural language, code, and more β€” trained on vast datasets with no task-specific programming required.

πŸ’¬ Chatbots✍️ Content generationπŸ”„ TranslationπŸ’» Code generationπŸ“Š Data analysis

Traditional AI

Rule-based systems and narrow ML models designed for specific, predefined tasks with structured inputs and outputs.

πŸ“§ Spam detection🎯 Recommendations🏭 Predictive maintenanceπŸ” Image recognitionπŸ“ˆ Demand forecasting
Dimension
Generative AI
Traditional AI
Setup
Prompt or fine-tune a pre-trained model
Define rules or collect labeled training data
Flexibility
Handles many tasks with the same model
One model per task
Data requirements
Few-shot or zero-shot β€” minimal examples needed
Large labeled datasets required
Output type
Open-ended text, code, structured data
Predefined categories or numeric predictions
Best for
Language, reasoning, creative and open-ended tasks
Narrow, well-defined classification or regression tasks

Key insight

LLMs don't replace traditional AI β€” they complement it. Use LLMs for language and reasoning; use traditional ML for high-precision, high-volume structured predictions.

A short timeline

Key milestones from early NLP to modern LLMs.