In my AI talks, I often reference OpenAI CTO Mira Murati’s 5 steps toward AGI. I like her framing, not just because it emphasizes thinking and reasoning, but because it anchors AGI in action.

AGI isn’t just about answering questions. It’s AI that can manage tasks, run systems, and ultimately operate and orchestrate an organization, made up of both humans and machines.

That’s the vision. But defining what counts as AGI? That’s still a moving target.

Even the leaders at the top AI labs can’t agree on what the finish line looks like. Is it general reasoning across any domain? Is it autonomy? Is it the ability to pursue goals without human prompting?

Ars Technica has a great deep dive on that exact question: What actually is this thing we’re racing toward, and how will we know when we’ve built it?

What is AGI? Nobody agrees, and it’s tearing Microsoft and OpenAI apart.

Several definitions make measuring “human-level” AI an exercise in moving goalposts.