Generative AI #

Generative Artificial Intelligence (GenAI) refers to a class of AI systems that can generate new content such as text, images, audio, video, or code, rather than only making predictions or classifications.

GenAI systems learn patterns and representations from large datasets and use them to produce novel outputs that resemble the data they were trained on.

How Generative AI Differs from Traditional AI #

Core Idea of Generative AI #

Instead of learning “what label to assign”, Generative AI learns “how data is structured” and then creates new data following that structure.

High-Level View #

flowchart LR
    DATA[Large Datasets]
    FM[Foundation Models]
    GEN[Generated Content]

    DATA -->|Training| FM
    FM -->|Generation| GEN

Key Characteristics of Generative AI #

• Trained on large-scale, diverse datasets
• Uses deep learning models, especially transformers
• Produces probabilistic outputs
• Can be adapted to many tasks using:
  • Prompting
  • Fine-tuning
  • Retrieval-Augmented Generation (RAG)

Types of Generative AI Models #

• Text generation
  (e.g. chatbots, summarisation, translation)

• Image generation
  (e.g. image synthesis, style transfer)

• Audio generation
  (e.g. speech synthesis, music generation)

• Code generation
  (e.g. programming assistants)

• Multimodal generation
  (combining text, images, audio)

Relationship to Other AI Concepts #

• Built on Deep Learning
• Implemented using Foundation Models
• Includes Large Language Models (LLMs) as a major subset
• Often combined with AI Agents for goal-driven behaviour

Where Generative AI Is Used #

• Conversational assistants
• Content creation
• Search and information retrieval
• Education and tutoring
• Software development support

Learning Path Within This Section #

• Foundation Models – the base models behind GenAI
• LLM – Model – language-focused generative models
• Transformer – the core architecture enabling GenAI
• AI Agents – systems that use GenAI for reasoning and action

Generative AI focuses on creating new content by learning the underlying structure of data, enabled by large foundation models and deep learning.

Home | Artificial Intelligence