By Tech Brew Staff
less than 3 min read
Definition:
Artificial generalized intelligence (AGI) is a hypothetical state in which a machine can learn and perform any task that a human can with equal or greater skill. It stands in contrast to artificial narrow intelligence (ANI), sometimes called weak AI, which can only perform a few specific tasks. Artificial super intelligence (ASI) is another hypothetical state in which an AI greatly surpasses our abilities.
Science fiction authors have been theorizing about AGI and ASI for decades, but researchers generally agree that existing AI models don’t meet the bar for AGI. There’s plenty of debate around when they might clear it, however. It’s a long-running conversation in the AI research community that’s become less fanciful and speculative as LLMs have gotten more and more advanced. Now, especially with the recent advent of improved reasoning models, a growing chorus in the tech industry and elsewhere is claiming that AGI may be upon us as soon as the next few years, while other experts dismiss this timeline.
Testing, testing
There’s no officially defined rubric for determining when an AI will have reached AGI, but there are some benchmarks that offer tests for determining it. One of the latest, the nonprofit Arc Foundation’s ARC-AGI-2, released in March 2025, still shows that a human panel outperformed all models on its tests—by a long shot.
The history of AI research has been marked by periods of inflated expectations around what AI can do, followed by stretches of disinterest and decreased funding when it fails to meet them—aka AI winters. While some have speculated about whether the current hype cycle might give way to an AI winter, recent progress hasn’t shown much sign of slowing down yet.
We’ll update this entry if AGI arrives (or maybe an AI will).
