How Accurate is Undetectable AI’s Detector? We Tested It On Itself

Published on:

Undetectable AI has change into fashionable in tutorial and running a blog circles as a strategy to circumvent false AI detection. It is a fast and simple strategy to safeguard your self towards false positives which have already claimed the status of multiple scholar.

Nevertheless, this is an attention-grabbing dilemma: If it is meant to cover AI writing utilizing AI, how can you make sure that its outputs will not be flagged as machine-generated? That is the rationale why we should check its output towards one of the best AI detectors out there.

Now, we have examined Undetectable AI towards different detectors earlier than. However what about its personal detector? Will it cross its personal check? Time to seek out out.

- Advertisement -

A Brief Dialogue on Undetectable AI’s Detector

We’ve had numerous Undetectable AI articles earlier than like this one, and this one, and this one, oh and right here’s one other one. The listing goes on. One characteristic that we’ve but to spotlight is its AI detector, or to be extra particular, detectors.

Undetectable AI doesn’t actually have its personal detector. As an alternative, it aggregates the AI probability scores from eight totally different detectors, particularly:

Undetectable AI additionally guarantees (and delivers) free re-rewrite if their output will get tagged as AI by a number of of these eight.

When you’re to study extra concerning the accuracy of every particular person detector, I extremely advocate studying our article the place we’ve examined most of their accuracy. I do know it’s a cliché line, however I promise, the reply would possibly shock you. 

- Advertisement -
See also  What generative AI can do for sysadmins

Undetectable AI vs. Its Personal Detector

So, sufficient foreplay, let’s get into the nitty-gritty. For this testing, we’ll ask ChatGPT to write down seven distinctive texts after which enter it into Undetectable AI. As soon as it’s generated the output, we’ll test if Undetectable AI’s detectors flag their very own output as AI.

Let’s get began.

Take a look at #1

Authentic GPT Textual content

Tweaked Output from Undetectable AI

Undetectable AI Detection Rating

Did it cross? Sure!
AI Detector Rating: 75% (6 out of 8)

Take a look at #2

Authentic GPT Textual content

Tweaked Output from Undetectable AI

Undetectable AI Detection Rating

Did it cross? Sure!
AI Detector Rating: 100% (8 out of 8)

Take a look at #3

Authentic GPT Textual content

Tweaked Output from Undetectable AI

Undetectable AI Detection Rating

Did it cross? Sure!
AI Detector Rating: 100% (8 out of 8)

Take a look at #4

Authentic GPT Textual content

Tweaked Output from Undetectable AI

Undetectable AI Detection Rating

Did it cross? Sure!
AI Detector Rating: 100% (8 out of 8)

Take a look at #5

Authentic GPT Textual content

Tweaked Output from Undetectable AI

Undetectable AI Detection Rating

Did it cross? Sure!
AI Detector Rating: 100% (8 out of 8)

- Advertisement -

Take a look at #6

Authentic GPT Textual content

Tweaked Output from Undetectable AI

Undetectable AI Detection Rating

Did it cross? Sure!
AI Detector Rating: 100% (8 out of 8)

Take a look at #7

Authentic GPT Textual content

Tweaked Output from Undetectable AI

Undetectable AI Detection Rating

Did it cross? Sure!
AI Detector Rating: 100% (8 out of 8)

See also  AlphaFold 3 Will Change the Biological World and Drug Discovery

Outcomes Tally

What Have We Realized? 

If we’re to take these scores at its face worth, it is protected to say that Undetectable AI’s tweaking is extremely efficient towards detectors. Nevertheless, you must all the time test their output additional as they use grammar errors to imitate human writing.

And in addition, there’s one other elephant within the room that I’ve but to say. Scroll again up and take a look at the eight detectors that Undetectable AI use. Discover one thing misplaced?

When you mentioned “OpenAI,” you would be right. OpenAI discontinued their AI detector months in the past and have famously regarded its detection as inaccurate. I personally do not know the place Undetectable AI sources its OpenAI probability rating. However because it’s all the time outputting “human,” then it is protected to imagine that it simply defaults to human whatever the enter.

That is really simply one of many inconsistencies I discover with this detector. One other is that once I enter Undetectable AI’s output to one in all their detectors’ official web sites, it typically offers me a distinct outcome from what I see within the Undetectable UI.

Past that, we are able to conclude that Undetectable AI does remarkably properly even towards its personal detector. When you’re focused on studying extra about this device, you may learn extra about it in our official assessment. Good luck!

- Advertisment -

Related

- Advertisment -

Leave a Reply

Please enter your comment!
Please enter your name here