# Kinds of AI — Artificial Intelligence

Source: https://www.skillbyai.com/en/artificial-intelligence/i-kinds

> Narrow and general, symbolic and learned.

## Useful distinctions

**Narrow AI** performs specific tasks (recognising faces, recommending films, translating text) and is what exists today, even when a single model handles many tasks. **Artificial general intelligence (AGI)** refers to hypothetical systems with broad human-level competence; when or whether it arrives is debated. Methods split roughly into **symbolic AI** (explicit knowledge and reasoning, transparent but brittle) and **statistical / learning-based AI** (patterns from data, flexible but harder to explain). Modern systems often combine them: a language model with tools, retrieval and search.

## Approaches compared

Strengths and weaknesses.

```text
approach         knowledge comes from     strengths                    weaknesses
search/planning  problem model            optimal plans, guarantees    needs a model, can be slow
logic/rules      experts write rules      transparent, auditable       brittle, costly to maintain
probabilistic    models + data            handles uncertainty          needs structure and data
machine learning examples                 flexible, scales with data   opaque, data-hungry, biased data
hybrid (LLM+tools) pretraining + tools     broad, adaptable             errors, cost, evaluation hard
```

## Use the simplest method that works

A rule or a search algorithm is often more reliable and explainable than a learned model for well-defined problems.

**Quiz:** What kind of AI exists in deployed systems today?

- [ ] AI with no data or rules
- [ ] Proven artificial general intelligence
- [ ] Conscious machines
- [x] Narrow AI focused on specific tasks

*Answer:* Narrow AI focused on specific tasks. Even broad models are evaluated task by task.