Artificial General Intelligence(AGI)
A hypothetical future AI that could handle almost any intellectual task a person can, rather than one narrow job. Think of the difference between a single-purpose tool and a capable colleague who can turn their hand to anything. In AI, this is called Artificial General Intelligence, or AGI. It does not exist today.
Artificial General Intelligence describes an AI that could learn and reason across any subject at roughly the level of a capable adult. It would not need to be trained separately for each new task. It could pick up something unfamiliar, work it out, and apply what it learned elsewhere, much as a person does.
This is worth contrasting with the AI we actually have. Today’s tools are what people call “narrow” AI. A model might write fluently, describe an image, or play chess at a superhuman level, but each system is built for its particular range of tasks. Even a broad tool like ChatGPT is predicting text rather than understanding the world the way a person does.
AGI vs the AI you use today
| The AI you use today | AGI (hypothetical) | |
|---|---|---|
| Range of tasks | Strong within the range it was built and trained for | Could turn its hand to almost any intellectual task |
| Learning something new | Needs retraining, or careful instructions from you | Could work out an unfamiliar task on its own |
| How it works | Spots patterns in its training data, for example predicting the next words | Would reason and understand across subjects, much as a person does |
| Reliability | Can sound confident while getting things wrong | Would need to be as dependable as a capable adult |
| Does it exist? | Yes, in tools like chatbots and image generators | No, it remains a research goal |
You will hear AGI discussed a great deal, sometimes as an exciting goal and sometimes as a worry. It matters to you mainly as a way of reading the headlines with a clearer eye. When a company says its latest model is “a step towards AGI,” that is a claim about ambition, not a description of what the tool can do for you today.
There is genuine disagreement among experts about whether AGI is close, distant, or even possible in the way it is usually imagined. Part of the reason is that they do not agree on a definition, so there is no agreed test that would tell us it had arrived. They also differ on whether making today’s large language models bigger will get there, or whether it needs ideas nobody has had yet. Nobody can tell you a date with confidence. For everyday purposes, it helps to treat AGI as a live research question rather than something you need to plan around right now.