Software Development

I Argued With AI Over Its Code — It Felt Shockingly Human, Until It Finally Conceded

AI Might Be Awesome In Coding, But It Can Be Wrong While Being Confident.

Photo by Adi Goldstein on Unsplash

This isn’t the usual case where I use AI to code, find a problem, point it out, and the AI admits the mistake and tries to fix it.

Instead, this time I suspected something was wrong and raised it with the AI. Rather than the typical “Oh, you’re right” response, it subtly implied I was mistaken, saying something along the lines of:

You’re thinking about this correctly, but …. …. it might not give you exactly what you want.

That reply chipped away at my confidence, even though the explanation just didn’t feel right.

The whole saga is almost like I’m working with a super intelligent coder that is knows a lot, and when questioned, don’t really bother to verify, but just explained through to prove one self correct.

With that suspicion in mind, I dug deeper, did thorough research, and eventually proved that I was right. My admiration for AI coding was brought down to a reality check. It’s impressive, but not entirely trustworthy.