What Is AI?
A complete, neutral guide to artificial intelligence — what it is, how it works, the types of AI, and the services that exist today.
§ 01 — Definition
What is AI?
Artificial Intelligence (AI) refers to computer systems designed to perform tasks that traditionally require human intelligence. These tasks include understanding language, recognising images, solving problems, writing software, analysing information, generating creative content, and assisting with decision-making.
AI does not think the way humans do. It does not possess consciousness, intentions, or creativity. Instead, it predicts patterns from data — patterns learned during training on vast amounts of text, images, and other information.
What AI is not
- — AI is not conscious. It has no self-awareness or feelings.
- — AI is not magic. It is mathematics applied at scale.
- — AI does not "think." It predicts the most likely output based on patterns.
- — AI does not assume responsibility. Humans do.
The carpenter analogy
Consider a professional carpenter. A carpenter may own the finest saws, drills, and machinery available. These tools increase efficiency and improve the quality of work, but none of them can design furniture, understand a client's vision, or make sound engineering decisions. The value lies with the craftsperson, not the tools. AI should be understood in exactly the same way.
§ 02 — History
A brief history of AI
Although the public became aware of AI only recently, the foundations were laid over seven decades. The development of AI has taken place in several distinct phases.
The idea of creating machines capable of intelligent behaviour emerged. Researchers like Alan Turing and John McCarthy began asking whether computers could imitate aspects of human reasoning.
§ 03 — Types
Types of AI
AI is classified in two ways: by capability (what it can do) and by function (how it operates). A critical fact: only Narrow AI exists today.
By capability
Narrow AI
AI designed for a specific task or domain. All AI that exists today — ChatGPT, Claude, self-driving, recommendation systems — is Narrow AI.
General AI (AGI)
AI that can perform any intellectual task a human can. Does not exist today.
Superintelligence (ASI)
AI that surpasses human intelligence across all domains. Does not exist yet.
By function
Reactive
Responds to inputs without memory of past interactions. Example: IBM Deep Blue (chess).
Limited Memory
Uses recent history to make decisions. Example: self-driving cars observing traffic.
Theory of Mind
Understands emotions, beliefs, and intentions of others. Does not exist yet.
Self-Aware
Has its own consciousness and self-awareness. Does not exist yet.
§ 04 — Technologies
The technologies behind AI
AI is not one technology — it is a lineage. Each layer is a subset of the one before it. Understanding this hierarchy is essential: machine learning is not the same as AI, and deep learning is not the same as machine learning.
Artificial Intelligence
The broad field: computers performing tasks that require human intelligence — understanding language, recognising images, solving problems, assisting decisions.
Example
Chatbots, recommendation systems, image recognition
Machine Learning
A subset of AI: computers learn patterns from data instead of being explicitly programmed with rules.
Example
Spam filters, fraud detection, speech recognition
Deep Learning
A subset of ML: neural networks with many layers learn from large datasets.
Example
Image generation, language translation, voice assistants
Transformers
A 2017 architecture breakthrough that processes language in parallel, enabling modern LLMs.
Example
The foundation of GPT, Claude, and Gemini
Large Language Models
AI systems trained on vast text to understand and generate human language by predicting the next word.
Example
ChatGPT, Claude, Gemini, Grok
§ 05 — Output Types
What AI produces
AI systems produce different types of output depending on their design. Understanding these categories helps you choose the right AI for the right task.
Generative AI
Creates new content — text, images, audio, video, or code — from patterns learned during training.
Examples: ChatGPT, DALL·E, Midjourney, Sora
Predictive AI
Analyzes data to forecast outcomes, classify items, or estimate probabilities.
Examples: Fraud detection, demand forecasting, medical diagnosis
Conversational AI
Understands and responds to human language in real-time dialogue.
Examples: ChatGPT, Claude, customer service chatbots
Agentic AI
Takes autonomous actions toward a goal using tools, data, and reasoning loops.
Examples: AI agents using MCP, agentic RAG, multi-agent systems
§ 06 — Agentic AI
Agentic AI
An AI agent is a system that can autonomously take actions toward a goal — perceiving, thinking, acting, and observing the results in a loop. Agentic AI is the newest and fastest-evolving category. These six terms define how it works.
Model Context Protocol (MCP)
The USB-C for AI
An open standard that lets AI agents connect to external data sources and tools — databases, APIs, file systems — through a unified interface.
Agent Loops
Perceive → Think → Act → Observe → Repeat
The cyclical process an AI agent follows: it perceives input, reasons about what to do, takes an action, observes the result, and repeats until the goal is met.
Skills
The tools an agent can use
The external capabilities an AI agent can call upon — searching databases, calling APIs, running code, reading files — to accomplish tasks beyond text generation.
Single vs Multi-Agent
One mind or many
A single-agent system uses one AI to handle a task end-to-end. A multi-agent system divides the work among specialised agents that collaborate or debate.
Agentic RAG
Retrieval with validation
An AI retrieval method where the agent actively searches, evaluates, and re-retrieves information rather than passively returning whatever a single search finds.
Agent Memory
Short-term vs long-term
Short-term memory is the context window — what the agent holds during a task. Long-term memory persists across sessions, allowing the agent to recall prior interactions.
§ 07 — The Landscape
The AI landscape
The AI market consists of hundreds of companies competing in different areas. No single AI is the best at everything. Instead of memorising company names, first understand the categories.
| Category | Products |
|---|---|
| General Purpose AI | ChatGPTClaudeGeminiGrokPerplexityDeepSeekMeta AIMS Copilot |
| AI Software Development | Base44LovableBolt |
| AI Image Generation | DALL·EMidjourneyStable DiffusionAdobe FireflyFLUX.1IdeogramRecraft |
| AI Video Generation | SoraRunwayKlingPika AILuma AIDescriptInvideo AI |
| AI Audio & Speech | ElevenLabsWhisperSuno |
| AI Coding Assistants | GitHub CopilotCursorTabnineReplitAskcodiCodigaQodo |
| AI Automation | ZapierMaken8nMonday.comWrike |
| AI Writing & Content | GrammarlyJasperCopy.aiQuillbotWritesonicRytrSudowrite |
| AI Knowledge Management | NotionMemTettra |
§ 08 — Cost
What AI costs
AI is priced in several ways:
Tokens
The smallest unit of text. API calls are billed per million tokens.
Credits
Platform-specific units for image generation, video, or compute.
Subscriptions
Monthly per-user fees for consumer and enterprise plans.
Compute time
GPU minutes for training or running custom models.
The biggest hidden cost of AI is not computation — it is poor thinking. Weak logic combined with AI produces poor outcomes more quickly.
§ 09 — How it works
How AI works
At its core, AI works in four steps:
Training data
AI is trained on vast amounts of text, images, or other data.
Pattern recognition
The system identifies statistical patterns in the training data.
Prediction
When given new input, the AI predicts the most likely output based on learned patterns.
Output
The predicted response is generated as text, image, audio, or action.
A Large Language Model specifically works by predicting the next word in a sequence. Given the text "The sky is," the model predicts "blue" as the most likely next word — not because it understands the sky, but because it has seen this pattern millions of times in training data.
§ 10 — Beyond the technology
All of the above are technologies. None of them decides how, when, or why a human should use them.
That is a separate question — the subject of AI Logic Engineering.
§ FAQ
Frequently Asked Questions About AI
01What is AI?+
AI refers to computer systems designed to perform tasks that traditionally require human intelligence — understanding language, recognising images, solving problems, writing software, analysing information, and assisting with decision-making.
02What is artificial intelligence in simple terms?+
In simple terms, AI is a computer system that can do things normally requiring human thinking — like understanding language, recognising patterns, and generating answers.
03What are the types of AI?+
AI is classified by capability (Narrow AI, General AI, Superintelligence) and by function (Reactive, Limited Memory, Theory of Mind, Self-Aware). Only Narrow AI exists today.
04What is the difference between AI and machine learning?+
Machine learning is a subset of AI. AI is the broad field of making computers do things that require human intelligence. Machine learning is a specific approach where computers learn patterns from data instead of being explicitly programmed.
05What is generative AI?+
Generative AI is a type of AI that creates new content — text, images, audio, video, or code — based on patterns learned from training data.
06What is an LLM?+
A Large Language Model (LLM) is an AI system trained on vast amounts of text to understand and generate human language by predicting the most appropriate next word or response.
07What is an AI agent?+
An AI agent is a system that can autonomously take actions toward a goal — perceiving, thinking, acting, and observing the results in a loop, often using external tools and data.
08What is AGI (artificial general intelligence)?+
AGI is a theoretical AI that can perform any intellectual task a human can. It does not exist today; all current AI is Narrow AI.
09Does AI exist today or is it theoretical?+
Only Narrow AI exists today. General AI (AGI) and Superintelligence (ASI) are theoretical and have not been achieved.
10How does AI work?+
AI works by learning patterns from large amounts of training data, then using those patterns to predict or generate outputs when given new inputs.
11Is AI conscious?+
No. Current AI does not have consciousness, self-awareness, feelings, or intentions. It predicts patterns from data.
12What is ChatGPT?+
ChatGPT is a conversational AI product made by OpenAI that uses a large language model to answer questions, write text, and assist with tasks.
13What is the difference between ChatGPT and Claude?+
ChatGPT is made by OpenAI and Claude is made by Anthropic. Both are large language models, but they are trained differently and have different strengths.
14How much does AI cost?+
AI costs vary: consumer subscriptions range from free to ~$20/month, enterprise plans cost more, and API usage is billed by tokens. The biggest hidden cost is poor thinking that leads to rework.
15What is MCP (Model Context Protocol)?+
MCP is an open standard that lets AI agents connect to external data sources and tools through a unified interface — described as "the USB-C for AI."
16What is agentic RAG?+
Agentic RAG is an AI retrieval method where an agent actively searches, validates, and re-retrieves information rather than passively returning whatever a single search finds.
17What is narrow AI?+
Narrow AI is AI designed for a specific task or domain. All AI that exists today — including ChatGPT, Claude, and self-driving systems — is Narrow AI.
18Can AI replace humans?+
AI can automate many tasks, but it cannot assume responsibility, make judgement calls, or define purpose. AI accelerates execution; humans remain responsible for decisions.
19What is AI used for?+
AI is used across nearly every industry — from virtual assistants and navigation to medical diagnosis, writing, and logistics. See our complete guide to the uses of AI.
20Who created AI?+
The concept of AI dates to the 1950s with pioneers like Alan Turing and John McCarthy. Modern AI is built by companies like OpenAI, Anthropic, Google, and Meta.