What is Generative AI?

Generative AI is a type of AI that can CREATE new content — text, images, video, code, audio — instead of only classifying or predicting.

How it differs from “traditional” AI

Type of AI	Example	Output
Discriminative	Spam email detection	A label (spam/not spam)
Predictive	House price forecasting	A number
Generative	ChatGPT, Midjourney	New text/images/video

Major categories of Generative AI (2026)

1. Text Generation

• LLMs: ChatGPT, Claude, Gemini, Llama
• Code: Claude Code, Cursor, GitHub Copilot
• Use cases: writing, translation, code, chatbots

2. Image Generation

• Diffusion Models: Midjourney, DALL-E, Stable Diffusion, FLUX
• Use cases: marketing, concept art, illustration

3. Video Generation

• Sora (OpenAI), Veo (Google), Kling
• Use cases: short-form video, ads, mockups

4. Audio / Music

• ElevenLabs: voice cloning, TTS
• Suno, Udio: original music
• Use cases: podcasts, voiceovers, jingles

5. 3D / Game assets

• Meshy, Spline AI: 3D models from text/images

Why did it explode after 2022?

• November 2022: ChatGPT launched — went viral, hitting 100M users in two months
• 2023: GPT-4, Claude, Midjourney v5 → quality crossed the “good enough for real work” threshold
• 2024-26: costs dropped 10-100×, open-source models caught up → applications everywhere

Economic impact

McKinsey estimates Generative AI could create $2.6-4.4 trillion in value per year (~the size of the UK’s GDP).

The most affected jobs:

• Marketing copywriters
• Software engineers (productivity up 30-50%)
• Customer support (chatbots handle tier 1)
• Designers (concepts and mockups)
• Translators

→ Less “replacement” and more “changing the way people work.”

Limitations

• Hallucination — making things up
• Bias from training data
• Unclear copyright status
• Energy intensive (each ChatGPT response ≈ the energy in one apple)

Related

• LLM
• Diffusion Model
• AI