What is AGI? (Artificial General Intelligence)

AGI is a hypothetical AI that matches or beats human performance across almost any intellectual task, not just the one it was built for.

How does AGI work?

Nobody has built one, so the definition is a test, not a design. The bar most researchers use is generalization: the system transfers a skill learned in one domain to a new problem it was never trained on, reasons under uncertainty, plans, and communicates, without a human rewriting it for each task.

Why does AGI matter?

For founders it matters mostly as a planning assumption. Every AI product bets on where capability lands in two years. Bet too low and a model update erases your feature. Bet too high and you ship a demo that cannot hold a customer today. AGI talk also moves capital, which moves valuations and hiring markets.

Where did AGI come from?

Mark Gubrud used the phrase in 1997. Shane Legg and Ben Goertzel popularized it around 2002 to separate general systems from the narrow, single task AI that dominated the field. Read the fuller history on Wikipedia.

How do you talk about AGI well?

Name the capability, not the label. Say what the system must do, on which inputs, at what accuracy, and who checks it. Avoid pitching AGI as a milestone your company will reach. Investors hear a timeline you cannot defend, and customers hear a promise you cannot ship.

Bottom line: AGI is a moving target with no agreed test, so build for the capability you can measure this quarter.

For more startup terminology, visit startupdefinitions.com.