AGI’s Paradox: When Perfect Answers Aren’t Enough

How AI Advances Are Shifting the Burden to Human Judgment

The internet is buzzing with yesterday’s AGI milestone. Let’s talk about it!

What is artificial general intelligence (AGI)? It’s an AI system with the capacity to learn, reason, and act effectively* across the full spectrum of cognitive tasks that humans can perform, without task-specific re-engineering.

OpenAI’s ARC-AGI performance

We’re not there yet, though yesterday OpenAI’s model o3 made a stir by demonstrating unprecedented performance on the Abstract and Reasoning Corpus for Artificial General Intelligence (ARC-AGI) benchmark.

Here’s an example of the ARC-AGI task from ARC-AGI… can you solve it? A task like this costs over $3,000 for model o3 to beat, but the progress is nonetheless real.

Every time a shiny new AI capability like model o3 shows up, the internet gets noisy with the usual gaggle of hypebeasts and curmudgeons weighing in on how good it actually is (or isn’t). But I’d prefer us to skip right to the logical conclusion of every AI release, asking:

“Imagine if AI was so good that you could get an instant answer to any question you wanted to ask. Or instant output for any request you made…