Can chatgpt actually prove the irrationality of square root of 2 or does it just reproduce known proofs
I asked ChatGPT to prove is irrational and it gave a perfect proof:
Assume in lowest terms. Then so . So is even, hence is even. Write . Then so . So is even, hence is even. Contradiction since and were in lowest terms.
But heres the thing: this is just Euclids proof from 300 BC. The AI didnt "discover" anything, it just reproduced what it was trained on.
My question: Is AI capable of generating NOVEL mathematical proofs? Or is it just a sophisticated search engine that regurgitates existing mathematics?
What would it take for an AI to genuinely extend mathematical knowledge?
ChatGPT is honestly pretty good at generating proofs that LOOK convincing but are subtly wrong. its a pattern matcher, not a mathematician.
but it got better! GPT-4 got the integral right when I checked.
Login to comment
1 Answer
I would solve this by writing the assumptions and the target on separate lines first. That usually reveals which theorem is actually needed.
A good structure is:
- state the definitions involved;
- transform the expression without skipping algebra;
- check edge cases such as zero, negative values, or boundary points;
- substitute the result back into the original question.
This may feel slower, but it prevents the most common math-answer problem: getting a plausible expression that does not actually satisfy the original conditions.
Could you add one more line on the condition where this method fails?
Login to comment