Picture this: 2,400 years after Socrates grilled his pupil on doubling a square, we're now pitting ancient geometry against modern AI. Researchers at Cambridge and Hebrew University fed ChatGPT the classic conundrum, and it nailed the square but flopped on the rectangle, confidently declaring no geometric fix exists—when one does. Hilarious, right? It's like watching a clever kid improvise a wrong answer in class, all earnest and off-base. But here's the intriguing twist: this isn't rote regurgitation; ChatGPT's riffing off prior chats, much like how we humans bootstrap from experience. The boffins suggest it's tapping into a 'zone of proximal development'—think of it as that Goldilocks zone where you're stretched just enough to grow, but not snapping under pressure.
As a techno-journalist who's seen AI hype cycles come and go, I love how this peels back the curtain on the black box. Sure, it's not 'thinking' like us—more like a probabilistic parrot with a PhD in pattern-matching—but calling its flubs 'learner-like' sparks real innovation potential. Imagine classrooms where kids prompt AI as a brainstorming buddy, not an oracle: 'Let's puzzle this out together' beats 'Spit out the solution.' It's pragmatic gold for educators, forcing us to teach verification skills over blind trust. No rose-tinted glasses here; AI proofs can be as dodgy as a politician's promise, so embedding critical eval in math curricula? Essential. Yet, the upside? Pairing ChatGPT with visual tools could turn dry theorems into interactive playgrounds, making math less 'eek' and more 'aha.' Let's lean in, test these digital apprentices rigorously, and build smarter hybrids—human curiosity plus machine muscle. Who knows, we might just double the fun in learning. Source: Scientists asked ChatGPT to solve a math problem from more than 2,000 years ago — how it answered it surprised…
