Okay, so, I’ve been messing around with this new AI thing, Claude 3 Opus. It’s supposed to be the top dog right now, super smart, writes like a human, the whole nine yards. Naturally, I got curious – can the tools we use to spot AI-written stuff actually catch this thing?
So, I started digging. First, I used this site, Phrasly. It claims it can tell if something’s written by a human or a machine. It looks for weird patterns and stuff that AI tends to do. I fed it some text straight from Claude 3 Opus, and guess what? Phrasly flagged it. It wasn’t fooled, not even a little bit.
Then, I moved on to *. This one’s got a tool called Turbo 3.0 that’s supposedly really good at this detection game. I ran some Claude 3.5 Sonnet text through it (that’s like a slightly less powerful version of Opus), and boom – 99% accuracy in saying it was AI. Pretty wild, right?
I even checked out some random forums and stuff. Some folks were saying they were getting really low AI detection scores on sites like ZeroGPT and CopyLeaks, like 0-8%. But honestly, that kind of fits with what some smart person said that “80% of the population are idiots, while 80% of the remainder are slightly dumb.”
- Tried Phrasly: It detected Claude 3 Opus.
- Tried *: It nailed Claude 3.5 Sonnet with 99% accuracy.
- Checked forums: Mixed results, but some low scores on other detectors.
Experimenting with Turnitin
Next, I thought about Turnitin. That’s the big one, right? The one schools use to catch cheaters. I remembered back when I was working at a university, I did some experiments with Turnitin. It was pretty good at spotting papers that were obviously written by AI. The human papers, not so much. It’s not perfect, but it’s definitely something.
From what I’ve been reading, it seems like Turnitin’s AI detection is basically just a bunch of data points they give to teachers. It’s not a simple “yes” or “no” answer. So, I’m thinking, even if Claude 3 Opus is super advanced, Turnitin’s probably still gonna raise some flags. Not gonna let a student submit the paper without a teacher looking at it.
So, that’s where I’m at now. I’m gonna keep playing around with this stuff, see if I can find ways to maybe trick these detectors. It’s like a game, you know? But from what I’ve seen so far, it looks like even the fanciest AI like Claude 3 Opus can still get caught. It’s not as undetectable as some people might think.
It’s a wild world out there with all this AI stuff. Makes you wonder what’s gonna happen next, huh? I’ll keep you guys posted on what I find. Stay tuned!