ChatGPT-4 can't solve Plato's classic math puzzle
A new University of Cambridge study found that ChatGPT-4 struggles with Plato's classic "doubling the square" problem—a math challenge that's been around for over two millennia.
Researchers wanted to see how the AI reasons through such old-school puzzles.
AI's answers were all over the place
Instead of using Plato's geometric approach, ChatGPT-4 went straight for modern algebra.
When researchers tried to trip it up or asked about rectangles and triangles, its answers were all over the place—sometimes missing key details or making mistakes.
With some extra nudging, though, it eventually got things right.
AI can sound smart, but it's just pattern-matching
The team noted that there are problems ChatGPT-4 can only solve with help.
The takeaway? While AI can sound smart, it relies on patterns and prompts rather than real understanding.
The researchers hope this reminds everyone (especially students) to double-check what AI says and think critically about its answers.