We've been thinking about AI wrong. The instinct is to compare it directly to human intelligence, like it's climbing some ladder toward our capabilities. But that's not how it actually works.
The concept of 'jagged intelligence' captures what AI really does. It can crush certain complex tasks while completely fumbling things a child could handle. There's no smooth progression, just peaks and valleys of capability that don't match our intuitions.
This matters because it changes how we should think about job displacement. Instead of AI replacing jobs from 'simple' to 'complex' in some orderly fashion, it's going to be messier and less predictable.
What AI does well today gives us clues about where it'll have impact. But the jagged nature means you can't just extrapolate in a straight line. A tool that writes marketing copy brilliantly might still need human help with basic context that seems obvious to us.
For anyone using AI tools at work, this reframes the question. It's not 'will AI replace my job?' but 'which specific parts of my work match AI's jagged strengths?' That's a more useful and honest way to think about it.
The debate shifts from existential worry to practical assessment. Understanding where those peaks and valleys are in your field helps you figure out where to lean on AI and where you're still essential.
