Answer
Types of AI
AI can be classified in multiple ways — by capability, by technique, and by application domain.
Classification by Capability
| Type | Description | Status |
|---|---|---|
| Narrow AI (ANI) | Excels at one specific task | Exists today |
| General AI (AGI) | Human-level intelligence across all tasks | Not yet achieved |
| Superintelligent AI (ASI) | Surpasses human intelligence in every domain | Theoretical |
Most AI today is Narrow AI — even ChatGPT and Claude are highly sophisticated narrow AI systems.
Classification by Technique
| Type | Description | Examples |
|---|---|---|
| Machine Learning | Learns patterns from data | Scikit-learn, XGBoost |
| Deep Learning | Neural networks with many layers | CNNs, Transformers |
| Generative AI | Creates new content (text, image, code) | GPT-4, DALL-E, Stable Diffusion |
| Reinforcement Learning | Learns via trial and reward | AlphaGo, robotics |
| Expert Systems | Rule-based reasoning | Medical diagnosis tools |
Classification by Output Type
| Type | Output | Examples |
|---|---|---|
| Language AI | Text / code | GPT-4, Claude, Gemini |
| Vision AI | Images / video | DALL-E 3, Stable Diffusion, Sora |
| Audio AI | Speech / music | Whisper, ElevenLabs, Suno |
| Multimodal AI | Text + image + audio | GPT-4o, Gemini 1.5, Claude 3 |
| Agentic AI | Actions in the world | AutoGPT, Claude Computer Use |
AI by Learning Approach
| Approach | How It Learns | Use Case |
|---|---|---|
| Supervised | Labeled examples | Classification, regression |
| Unsupervised | Unlabeled data, finds structure | Clustering, anomaly detection |
| Semi-supervised | Mix of labeled + unlabeled | NLP pre-training |
| Self-supervised | Creates its own labels | LLM pre-training (predict next token) |
| Reinforcement | Rewards and penalties | Games, robotics, RLHF |
Current Dominant AI Technologies (2025)
- Large Language Models (LLMs) — GPT-4, Claude 3.5, Gemini 1.5
- Diffusion Models — DALL-E 3, Stable Diffusion, Midjourney
- Multimodal Models — GPT-4o, Gemini 1.5 Pro
- Reasoning Models — o1, o3, DeepSeek-R1
- AI Agents — autonomous task-completing systems
Key Takeaway
The AI landscape is broad. For Gen AI engineering, the most relevant types are:
- LLMs for text/code generation
- Diffusion models for image generation
- Multimodal models for mixed inputs
- Agentic AI for autonomous workflows