The term AGI is often used without a clear definition, leading to unproductive debates. A better approach is to define it functionally. Either AGI is achieved when AI's impact fundamentally transforms society, or it should be viewed as a spectrum of increasing generality, not an all-or-nothing milestone.

As soon as OpenAI's Astra model nearly 'solved' the ARC AGI 3 benchmark, its creator immediately 'moved the goalposts' by announcing the next version will focus on 'open-ended invention.' This shows how the very definition of AGI is a moving target, constantly being redefined by technological breakthroughs.

A practical definition of AGI is an AI that operates autonomously and persistently without continuous human intervention. Like a child gaining independence, it would manage its own goals and learn over long periods—a capability far beyond today's models that require constant prompting to function.

Greg Brockman describes the imminent arrival of AGI not as a singular event where AI becomes uniformly superhuman, but as a 'jagged' reality. The AI will be superhuman at most intellectual computer-based tasks while still struggling with some basic tasks a human can do, making a clear definition difficult.

"Superintelligence" is clearly defined as AI that is better, faster, and cheaper than the best humans at everything. In contrast, "AGI" (Artificial General Intelligence) is a vague term for general-purpose AI, a milestone that current models have arguably already achieved.

Despite claims that GPT-6 Astra marks the arrival of AGI, its capabilities appear to be an incremental advance, potentially just catching up to competitors like Anthropic's Fable 5.1. This framing overlooks that AGI is a spectrum of capabilities, not a single binary event, suggesting the claim is primarily for marketing.

The definition of AGI is a moving goalpost. Scott Wu argues that today's AI meets the standards that would have been considered AGI a decade ago. As technology automates tasks, human work simply moves to a higher level of abstraction, making percentage-based definitions of AGI flawed.

Even as AI models surpass technical AGI benchmarks, the host argues people will keep moving the goalposts. The true, socially accepted definition of AGI will be its "feel"—its ability to generalize and execute complex, nuanced tasks with minimal instruction, like a human.

AGI isn't a single switch but a tiered system defined by capability and breadth. The Google DeepMind framework categorizes AGI into levels based on the percentage of humans an AI can outperform on a given task, moving from outperforming 50% (Tier 1) to 100% of humans.