There’s no single list of the types of artificial intelligence. AI can be classified by its capability (ANI, AGI and ASI), by the technology it uses, by how it learns, by what it produces and by how autonomously it acts. This guide explains each classification, the real difference between ANI vs AGI vs ASI and which types of AI actually exist today.
- How artificial intelligence is classified
- Types of AI by capability: ANI, AGI and ASI
- ANI vs AGI vs ASI compared
- Types of AI by technical approach
- Types of AI by how they learn
- Types of AI by what they produce
- Types of AI by autonomy
- Reactive machines, limited memory and theory of mind
- One system can belong to several categories
- Which types of AI exist today?
- Types of AI FAQ
- Keep reading
Quick answer: every AI system in use today is ANI (artificial narrow intelligence), even very capable assistants like ChatGPT, Gemini or Claude. AGI (artificial general intelligence) is a research goal with no agreed test, and ASI (artificial superintelligence) is a hypothetical scenario. Beyond that scale, the same system can be generative, multimodal, conversational and agentic at the same time.
How artificial intelligence is classified
| Criterion | Main categories | What it describes |
|---|---|---|
| Capability | ANI, AGI and ASI | How broad and general its abilities are |
| Technical approach | Rules, machine learning, deep learning | The technology used |
| Learning | Supervised, unsupervised, reinforcement, self-supervised | How it learns from data |
| Output | Predictive, generative, classifier, recommender, conversational, vision, robotics | What it produces |
| Autonomy | Assistants, agents, agentic systems | How independently it acts |
Types of AI by capability: ANI, AGI and ASI
ANI: artificial narrow intelligence
ANI, also called weak or narrow AI, covers systems designed for specific tasks or domains: a spam filter, a recommendation engine, a voice assistant, a chess program or a language model. They can be extremely good, sometimes better than humans, within their scope, but they don’t have flexible general intelligence across any task.
- Specialized in defined tasks or contexts
- Can outperform humans in narrow areas
- Depend on their training and design
- All AI in real use today falls here
AGI: artificial general intelligence
AGI describes a hypothetical system able to learn and perform a wide range of intellectual tasks at a level comparable to humans, adapting to new problems without being specifically built for them. There’s no universally accepted definition or test, and companies and researchers use the term in different ways.
AGI doesn’t necessarily mean consciousness. General capability and subjective experience are different things: a definition of AGI can focus on learning and adaptation without implying a machine is aware.
Does AGI exist today? There’s no consensus that any current system is AGI. Today’s leading models are broad and impressive, but their performance is uneven and they still have clear limitations.
ASI: artificial superintelligence
ASI refers to a hypothetical intelligence that would far exceed the best human abilities in practically every field, including science, strategy and creativity. It doesn’t exist, and it’s mostly discussed in the context of long-term AI safety and governance.
ANI vs AGI vs ASI compared
| Feature | ANI | AGI | ASI |
|---|---|---|---|
| Meaning | Narrow, specialized AI | General AI | Superintelligence |
| Status | Real systems in use | Research goal with no universal test | Hypothetical scenario |
| Scope | Defined tasks and contexts | A wide variety of domains | Would far exceed humans in nearly everything |
| Examples | ChatGPT, spam filters, recommenders, self-driving features | None agreed | None |
Types of AI by technical approach
- Rule-based and knowledge-based AI: follows rules written by experts, as in classic expert systems. Predictable and explainable, but rigid.
- Machine learning: learns relationships from data instead of hand-coded rules.
- Deep learning: a type of machine learning using neural networks with many layers; it powers image recognition, speech and generative AI.
Types of AI by how they learn
- Supervised learning: learns from labeled examples to predict labels for new data.
- Unsupervised learning: finds patterns and groups in unlabeled data.
- Reinforcement learning: learns by trial and error through rewards and penalties.
- Self-supervised learning: creates its own learning signal from the data, such as predicting missing words; it’s key to training large language models.
Types of AI by what they produce
- Predictive AI: estimates an outcome, value or probability, such as demand or risk.
- Generative AI: creates new content such as text, images, audio, video or code.
- Discriminative or classifier AI: sorts inputs into categories, such as spam detection or medical image triage.
- Recommendation systems: suggest products, videos or songs based on behavior.
- Conversational AI: chatbots and voice assistants that interact in natural language.
- Computer vision: recognizes and interprets images and video.
- Robotics and autonomous systems: perceive their surroundings and act in the physical world.
Types of AI by autonomy
- Assistants and copilots: help a person who stays in control, such as suggesting text or code.
- AI agents: receive a goal, plan actions, use tools and act with some autonomy.
- Agentic AI systems: coordinate several agents or processes, break goals into tasks and act over longer periods with less direct human input.
Reactive machines, limited memory and theory of mind
Another popular classification, used in many introductory courses, describes four levels:
- Reactive machines: respond to specific inputs without building memory of past experience, like classic chess engines.
- Limited memory: use recent data or context to make decisions; most current AI, including language models and self-driving features, fits here.
- Theory of mind: would understand emotions, beliefs and intentions of others. Still a research concept.
- Self-aware AI: would have consciousness of its own. Purely hypothetical.
One system can belong to several categories
These classifications aren’t mutually exclusive. A modern AI assistant is narrow AI (by capability), deep learning (by approach), trained with self-supervised and reinforcement learning, generative and conversational (by output), multimodal, and it can act as an agent when connected to tools.
Which types of AI exist today?
Specialized, predictive, generative, conversational, vision, robotic and agentic systems are all real and widely used. Every one of them is narrow AI. AGI remains a goal under debate, and ASI, theory-of-mind and self-aware AI remain hypothetical.
Types of AI FAQ
There’s no single number. AI can be classified by capability, technology, learning method, function, modality and autonomy.
ANI is specialized AI; AGI is a general capability comparable to humans across many domains; and ASI is a hypothetical intelligence far superior to humans.
AGI would match human-level ability across many tasks. ASI would far exceed the best humans in practically every field. Neither exists today.
Under the ANI/AGI/ASI scale it’s generally considered narrow AI. It can perform many tasks, but there’s no consensus or universal test that certifies it as AGI.
No. A model can generate content in many formats and still be specialized, narrow AI.
A system that can receive a goal, plan actions, use tools and act with some degree of autonomy.
Systems that coordinate agents or processes, split goals into tasks and carry out actions over longer periods with less direct human involvement.
No. General capability and consciousness are different concepts; AGI definitions usually focus on learning and adaptation.