With ChatGPT’s new GPT Store, you can now get tools set up for specific help. From making stories to giving law help, computer coding, and even medical facts, these special tools give more than the usual ChatGPT.
There are tools out there that say they can get around AI detection, but do they really work? Specifically, the GPTs that say they can dodge detection don’t hold up against specialist stealthy AI writers.
If we strive for genuine originality, even specialist stealthy AI writers or crafty methods can’t fool our sharp eyes. This proof comes straight from academic research.
With the opening of the GPT store, it’s no wonder that some bots marked “humanizers” are top sellers. These are engineered to outsmart rigid AI scripts and infuse the material with a more human touch.
But, can these “human-like” models honestly churn out AI material that can’t be found out?
We’ve put their claims to the test. Let’s see how they fare.
Our selection came from the most favoured AIs in the writing segment of the GPT shop, which includes ‘Humanize AI’ and ‘Humanizer Pro’
Firstly, for this speedy exam, we took five known AI writings from our detection standard dataset. Next, each text was run through ‘Humanize AI’, and its genuineness was checked by Netus.ai. We repeated this process with ‘Humanizer Pro’.
For the folks who spend their daily routine finding AI-created texts, the results weren’t unexpected.
We wanted to check how well it could spot text made by an AI. First, we fed it five samples all made by AI, as a basic test. As expected, it said each and every piece was 100% AI-made. Then, we used a tool called Humanize AI from the GPT Store. This tool tries to “humanize” AI-made writing. After doing this, we tested it again, and here’s what Netus.ai said:
After Humanize AI was used:
Sample 1 had 97% detected AI, while the rest, Samples 2-5 still had 100% detected AI.
We tried the same, using a different tool, Humanizer Pro from GPT Store. After getting it made more human-like, we checked with Netus.ai again. The results?
After Humanizer was used: It said every sample, Samples 1-5, was 100% produced by an AI
AI detectors use smart math rules and trained machine models to spot fake text changed by machines. These machines make the fake text look real, but the best detectors can spot the tiny clues that tell us if a machine or human wrote it.
One important tool is guess-checking. Machines like GPT-3 or GPT-4 write text by guessing the next word. This guessing is based on rules. Even when machines change the fake text to look real, their guesses follow rules. Detectors check these guesses against lots of real text written by humans and can spot anything strange.
Another way is by looking at the sentence structure and the meaning of the words. Machines often repeat the same sentence style and use regular grammar rules. Detectors look for repeated sentence lengths, mirrored phrases and formal words—which are often used in machine-written text. The best detectors can even check if the text makes sense in context, and spot text that lacks real depth or doesn’t flow well.
Word choice and how often words are used matters a lot too. Human writers use different words and unique phrases based on their own experiences. But machines often use the same phrases they’ve been fed. When there’s not much variety in the words, that’s a clue that a machine was involved.
Lastly, detectors can use special models to compare how much a piece of text is like known fakes. By matching the meaning of a text with millions of other checked samples, detectors can find fake text that’s been altered by machines.
AI writing and detection are about to evolve quickly as they get better and better. We’re going to see AI-created articles that are just like ones written by humans, while the tools we use to find them will keep upping their game.
Picture this: future AI writing systems will be way ahead of what we have right now, making articles that are just like a person wrote them. They can match the way a human talks, their creativity, and even their sense of humor. These AI’s will be able to sound just like a particular person or cultural group, making it tricky for detection tools to spot them. Some of them might even mix together writing with pictures or sound to make content more interesting.
AI detection systems will also get a lot smarter. They’ll be able to understand context much deeper than they do now. They might be able to compare a suspicious text with a whole bunch of other texts, follow real-time news, and understand how a story is being told. These systems might also understand natural human language better, which can help them figure out the aim and complexity of a text.
There might be a role for blockchain technology, too, in making sure content is original. Writers might use it to time-stamp their work and prove it’s theirs. That way, it’ll be harder for AI to pretend it wrote an article that was actually done by a human.
As AI starts to change how we make content, governments and big businesses might step in with new rules about what’s fair play. They might, for example, need AI-created content to be clearly labeled or have a sort of digital watermark. There may even be legal rules to make sure AI-created content is completely transparent.
Rather than taking over from writers, AI might become their buddies. Writers could use AI systems to make their writings better while still keeping their personal touch. In this kind of partnership, the best parts of human creativity will join forces with AI’s amazing efficiency to make lots of high-quality content.
AI is changing how we write and spot fake writing. Soon, AI machines will write articles just like humans, and the tools used to detect them will also improve.
Picture AI technology in the future outdoing current capabilities, crafting articles that seem like they’ve been penned by humans. They’ll echo human speech, creativity, and humor. They will imitate specific people or cultural groups which would be hard for detection tools to reveal, with some even merging writing, images and sounds for an engaging twist.
The AI detection technology will become highly advanced. They will understand the surroundings of a text much better. They may be able to compare a doubtful piece with others, monitor real-time updates and comprehend the narrative style. They will also grasp human language better, assisting in figuring out the text’s purpose and intricacy.
Blockchain technology could ensure that content remains unique. Authors may use it to time-stamp their output to affirm it as theirs, making it difficult for AI to falsely take credit for their work.
As AI reshapes content creation, regulations may be introduced by governments or large corporations to maintain fairness. For instance, explicit labeling or a digital mark might be required for AI-generated content. Strict laws could be enacted to make AI-created content entirely transparent.
Instead of replacing writers, AI might collaborate with them. Authors might utilize AI technology to enhance their writing while retaining their unique style. In such a collaboration, human creativity will couple with AI’s extraordinary efficiency to produce a plethora of high-quality content.
It’s okay to wish for a magic trick that can tell machine writing apart and make it feel more like a human wrote it. Sadly, our checks show that AI, even with unique GPTs, has some distance to cover before truly matching up to the human style. As these new “human-like” GPTs roll out, we’ll keep testing and updating you.
Right now, even the special GPTs from stores that claim to “add a human touch” to your words may fail when faced with true AI-detecting software. If you have plans to use a “human-touch” GPT to make your work indistinguishable, you might want to reconsider. This is because top-notch AI writing detectors can almost always identify such work, every single time.
Through digital leadership we empower people to leverage the opportunities in global economy
Through digital leadership we empower people to leverage the opportunities in global economy
@ 2024 Netus AI.
