The world’s biggest technology companies say they are racing to build artificial general intelligence. Here’s what AGI means, how it differs from today’s AI, and why it divides experts.
Most AI in use today is described as narrow. It is built for a specific job, such as recommending videos, recognising faces or translating text.
Artificial general intelligence, or AGI, is the idea of an AI system that can learn and perform almost any intellectual task a human can, from writing software to doing scientific research, and apply what it learns in one area to another.
There’s no single agreed definition. Some describe AGI as AI that can do most work people do on a computer as well as an expert. Others focus on whether it can learn new skills on its own, or set and pursue its own goals.
Why it’s suddenly a big deal
For decades, AGI was mostly a topic for science fiction and academic debate. That changed with the rapid progress of large language models such as ChatGPT, which can now write, reason through problems, code and pass difficult exams.
Companies including OpenAI, Google DeepMind, Anthropic and Meta say building advanced general AI is part of their mission. They are spending hundreds of billions of dollars on data centres and chips in the belief that more computing power and data will keep producing more capable systems.
The idea even appears in business contracts. Microsoft’s partnership with OpenAI has included terms that depend on whether AGI is achieved.
When could it arrive?
Opinions vary enormously. Some technology leaders say AI that matches humans at most tasks could arrive within a few years. Many researchers are more cautious, arguing today’s systems still make basic mistakes, can’t reliably plan over long periods and lack real understanding of the world.
Others argue the term is so vague it will be hard to say when it has been reached, and that progress will be gradual rather than a single breakthrough.
The promise and the risks
Supporters say AGI could speed up scientific discovery, develop new medicines, boost productivity and help solve problems such as climate change.
Critics and many AI researchers warn of risks, including large-scale job losses, misuse by criminals or hostile governments, the concentration of power in a few companies, and the challenge of making sure very capable systems act as humans intend.
Governments, including Australia’s, are working out how to regulate advanced AI, and international summits have brought countries and companies together to discuss safety.
