The Jagged Frontier
The most dangerous misconception about AI isn't that it'll replace us. It's that most of us have no idea where it actually excels and where it falls flat. We picture AI competence as a smooth gradient: easy tasks easy, hard tasks hard. The reality is what researchers call a "jagged frontier", an unpredictable mess of peaks and valleys that ignores everything our intuition tells us about intelligence.
You see it in odd places. The same system that writes a half-decent sonnet can fumble basic arithmetic. An AI that passes a legal exam might lose to you at noughts and crosses (tic-tac-toe, for my American readers). These contradictions force you to rethink what "capability" even means.
The creative side has proved the biggest surprise to me. When ChatGPT scored in the 99th percentile on the Torrance Test of Creative Thinking, it knocked a hole in the comfortable assumption that creativity is a uniquely human trick. There was another study I came across where humans and AI went head to head generating product ideas, and the AI apparently came out well ahead on the top-rated ones - I'd take the exact figure with a pinch of salt, but the direction of travel is hard to ignore. Maybe creativity is less about divine inspiration and more about systematically rummaging through the space of possibilities.
I recently finished "Co-Intelligence" by Ethan Mollick, where he draws a nice parallel with the Wright brothers. They succeeded where cleverer, better-funded people failed because they combined different worlds. Watching birds plus hands-on experience as bicycle mechanics got them off the ground. AI does something like that today, pulling together ideas from the whole sprawl of human knowledge. It doesn't replace human insight, it amplifies it - the comprehensive reference the Wright brothers never had.
Once that clicks, it changes how you work with these things. The traditional setup parks the human as quality control, a glorified spell-checker for the machine. Mollick's framing is richer, and frankly I prefer it. Instead of just catching errors, you become the conductor, steering an intelligent system towards a goal and bringing human judgement and values to the parts that need them. The relationship stops being tool-use and starts being collaboration.
AI throws up options and spots patterns at a scale no person can touch. Humans bring something else entirely. We notice the telling detail, we feel the emotional weight of a thing, we grasp the implications the model misses about our particular situation. The value isn't in racing the computer on raw horsepower; it's in the wisdom that turns all that computation into something worth having.
People worry, fairly, about AI flooding the world with mediocre slop. That worry misses the bigger prize. Treat AI as a production line and yes, you get more sludge. Treat it as a thinking partner and the whole thing shifts. The question stops being how much you can churn out and becomes how deeply you can dig into ideas that actually matter. We can all now poke at genuinely complex ideas in a way that simply wasn't possible before, and I'd put that somewhere alongside the arrival of the computer and the internet.
Some critics will say future AI won't need us in the loop at all, so why bother? Why care about collaboration when we're all about to be obsolete? That's a tidy way to talk yourself out of using the best tools you'll ever have. Sitting on your hands waiting for a hypothetical future where AI needs no human input just means you forfeit everything it can do for you right now. The jagged frontier is here, today, and it's not going anywhere soon.
There's a particular smugness in writing off AI's creative potential. It ignores that these systems have absorbed more human knowledge than any one person could read in a thousand lifetimes. We've always stood on the shoulders of books (I'm doing exactly that with "Co-Intelligence"), mentors, and the collective brain-dump of the internet. Now we have a different kind of intelligence to stand on too.
How AI reshapes work isn't really a "will it" question. It's a question of how we decide to engage with this jagged frontier. I'd rather be a collaborator with it than a rival scrapping with it for relevance.