Artificial General Intelligence
Artificial General Intelligence (AGI) is a hypothetical AI system with human-level competence across every cognitive domain: it could learn any new task from instructions or experience, transfer skills between unrelated fields, and plan long projects without being retrained or rebuilt for each one. No such system exists today; current models are general in what they accept but narrow in what they reliably do.
No jargon, no math
In plain words, what is it?
The word general is the whole story. Today's models are general at the front door: they accept any text, so they look universal. Behind the door they are narrow: they do one pass, hold no memory of yesterday, and accuracy drops sharply on multi-step problems they have never seen. AGI is the name we give to the system that would close that gap. Four capabilities would have to come together: transfer (skills learned in one domain working in another), continual learning (improving from experience without a full retrain), long-horizon planning (holding a goal through hundreds of dependent steps) and agency (acting in the world and correcting itself). Where it breaks: there is no agreed test. Benchmarks built to measure AGI get solved and stop measuring anything, and the definition keeps sliding to whatever machines still cannot do.
Terms that live nearby.
Artificial Intelligence
Artificial intelligence is software that performs tasks which normally need human judgement: understanding language, recognising images, planning, deciding. Modern AI is built by training models on data rather than hand-writing rules, which is why it copes with messy input a rigid program cannot.
ReadLarge Language Model
A large language model (LLM) is an AI system trained on massive collections of text to predict the next token, which gives it a general ability to write, summarise, translate and reason without being programmed for each task.
ReadAlignment
Alignment is the work of making an AI system do what you actually meant, not merely what you literally typed: its goals, limits and refusals are shaped during training and deployment so its behaviour stays within what operators and society intend.
Read§ 03
Want this running in your stack?
We design and ship AI systems with permissions, logs and human approval. It starts with a 30-minute scoping call.