If you’ve been paying attention to the tech world recently, you've probably noticed ChatGPT & GPT-4 and realized they're nothing short of amazing.
Over the last year we’ve seen a technical revolution of something similar to when the internet came out. OpenAI, the company behind it all, continues to ship groundbreaking products across the AI world & isn't showing any signs of slowing down. Other companies like Apple, Google, and Meta have also hopped on board.
Their most recent change, GPT-4, is one of the most powerful language model ever created. It's capable of producing writing with extreme precision & quality of human writers.
If this is true, has AI writing reached the point where it’s now completely undetectable and indistinguishable from human writing? Not quite…
The way AI Writing works is by predicting the following words of a sentence based on the words that come before it. Think about it this way:
If you made someone read 10,000 cardiology textbooks and then asked them to recite how the heart works, they’d probably have a boilerplate answer nailed into their head. I mean, they’ve been reading about the same processes at such an extreme scale.
This analogy is a greatly oversimplified version of how generative AI works. You train a model on a bunch of data, and when trying to retrieve knowledge, you’ll get very specific & trained answers.
Now when it comes to how much info ChatGPT/GPT4 is trained on, it hurts to even think about. Truly incomprehensible. It's not even public knowledge
GPT-3 (the previous model) was trained on 175 billion, so it's hard to imagine how much bigger this model actually is.
The same way AI tools generate writing is the exact opposite method of how it’s detected.
AI detection tools like Originality and Content at Scale try to reverse engineer the writing you enter to see how similar it would to an AI tool like ChatGPT rewriting it.
Here’s an example sentence: I really want to go to the ______. Maybe 90% of the time the AI will say “beach”, but that’s not what I was going to say, my sentence would’ve ended with “trampoline park”.
Human writing is very versatile and AI writing just tries to repeatedly emulate patterns. The same logic also applies for an entire article, essay, or report. The more likely writing aligns with what an AI would’ve written by itself, the higher chance it was written with AI.
So, what’s the solution?
You have to use words and structure sentences in a way that an AI writing generator wouldn’t use. How is this possible? I know, it's weird and frankly stupid.
Well if you want to be really technical, you can't definitively predict AI text. It's just not possible. There's no watermark that gets left behind to prove anything (unlike how plagiarism checkers show you the exact content somebody stole from).
With AI detection, it's just a bunch of words. But for now, AI still writes at a level that can be somewhat predicted, at least if the vocabulary and sentence scrambling isn't too complex. If not, it's pretty much impossible. As these AI tools get better, the detectors are going to be fighting a tough battle.
For now, you could combine a mix of human intuition & online tools that can help scramble your text in ways tools like ChatGPT couldn't write it... at least for now. Here's how:
How to Make AI Writing Undetectable & Bypass Detection
A few online tools make it fairly easy to remove obvious AI-written text from tools like ChatGPT, Jasper, and really any kind of AI writing software.
These Undetection tools all make use of a combination of machine learning algorithms and natural language processing techniques to identify patterns in the text that make it obvious you're looking at AI-generated content.
To make AI writing undetectable from these online detection tools, you need to focus on making the text sound more real, ensuring it flows logically, and making sure it doesn't always sound the same.
While it may be tempting to use AI to breeze through assignments or work tasks, it's important to be mindful of the potential consequences. Undetectable AI-generated content can lead to the spread of misinformation, plagiarism, and other harmful activities. Please don't do anything unethical or illegal.
Using AI Writing Scrambling Tools
AI has led to tools that hide AI-written text better. They shuffle words, rephrase sentences, and use synonyms. For example, "Exercise improves heart health" becomes "Heart health gets better with regular activity." This keeps the meaning but seems more human. It confuses detection tools by mimicking human writing's variety and style.
Undetectable.ai
The first tool is to use Undetectable.ai. They addresses the challenge of transforming AI-generated content that has been flagged by detection tools into high-quality writing that appears to be written by a human. This user-friendly platform is designed to help you change your content to suit various writing styles, such as essays, general writing, marketing material, and more.
To use Undetectable AI, simply paste your text into the tool, and choose the appropriate readability level and tone for your specific writing style.
Once you click the "Humanize" button, the service will work its magic, converting your AI-generated content into text that is not only coherent and engaging but also indistinguishable from human-written material.
I took a ChatGPT response about AI and ethical usage and rephrased it to a doctorate essay. It produced something very advanced and passed all the AI detectors (92% human from Content at Scale!)
By leveraging the power of Undetectable.ai, you can ensure your content maintains a consistent writing style while avoiding potential issues with AI-detection tools.
Make sure to reread content, as sometimes the rephrasing inserts synonyms and styles that don't actually make sense given the context of the writing. These tools should always be used as an assistant, not a replacement.
Also remember to use this service responsibly and ethically, keeping in mind the potential consequences of creating and submitting writing that was modified by a tool.
HideMyAI
What's funny is this tool just popped up out of nowhere and seems to be the best in terms of quality. While HideMyAI takes a little bit longer to process text & doesn't come with as much complexity, you're still able to get some amazing undetectable content in a matter of seconds. It works the same, just paste your writing into the box:
Ok, so this is obviously written with ChatGPT. Now let's paste it into HideMyAI:
New result reads very well. But how does it do when we put it back into the detector? Pretty damn impressive. Not a single trace.
Tweak Sentence Syntax
One of the most effective strategies for turning AI detectable writing into content that passes as human-written is to modify the sentence syntax. This involves adjusting the arrangement of words and phrases in a sentence, as well as the relationships between them.
By doing so, you can create subtle variations in the structure of your text, making it harder for AI detection tools to identify patterns that are characteristic of machine-generated content, similar to what QuillBot offers. Doing this manually is similar to what we spoke about earlier but requires a lot more personal fine tuning (a software won't do this as extensively as the human eye).
To tweak sentence syntax, you can experiment with changing the order of clauses or using different sentence constructions, such as switching between active or passive voice. For example, if you have a sentence like "The AI-generated content was detected by the tool," you can switch to an active voice and say, "The tool detected the AI-generated content."
Obviously you'd have to go more in-depth than that, but you can apply that idea to the entirety of your writing.
You should also be utilizing a diverse range of sentence lengths and structures. In natural human writing, it's common to find a mix of short and long sentences, as well as a blend of simple, compound, and complex structures.
By mixing this into your AI-generated content, you can mimic the rhythm and cadence of human writing to a waaaay larger degree.
Just remember to strike a balance between tweaking the sentence syntax and maintaining the original meaning and coherence of your content to ensure that it remains both engaging and informative for your readers.
Readability is key!!! Nobody wants to read poorly structured sentences.
Use Higher Level Vocabulary
Another great technique for making AI-generated content less detectable is to incorporate higher-level vocabulary into your writing. This not only elevates the quality of your content but also helps to obscure the telltale signs that it was produced by an AI.
To use higher-level vocabulary, consider replacing common or overused words with more precise or specialized terms. For instance, instead of using the word "important," you could opt for "crucial," "pivotal," or "indispensable."
Likewise, you can employ technical jargon or industry-specific terminology when discussing a particular subject, which lends credibility and depth to your content. Tools like QuillBot are actually perfect for this, they let you paragraph word by word & sentence by sentence until you're happy with your new paragraph.
Keep in mind that while using advanced vocabulary can make your writing more human-like, you need to find a balance between sophistication and readability (I'm assuming you'll have real humans read this somewhere, right?!)
Your content should remain accessible and engaging to your target audience, so avoid using obscure words that might confuse or alienate readers. Always keep in mind user intent, it wouldn't be worth over sophisticating your text if you are writing an article for 5th graders just for the sake of bypassing AI detection.
In addition to higher-level vocabulary, you can also make use of synonyms and paraphrasing to add richness to your content. This approach not only helps to mask the AI-generated nature of your text but also improves its overall flow and cohesion.
By diversifying your word choice and experimenting with different ways to express your ideas, you can create a more dynamic and compelling piece of writing that effectively evades detection by AI analysis tools while still naturally getting your point across.
Final Thoughts & Remarks
Transforming AI-generated content into text that is indistinguishable from human-written material is a fascinating and an increasingly relevant topic ever since ChatGPT launched. By refining the AI's training data, tweaking sentence syntax, and incorporating higher-level vocabulary, people can now create content with tools like ChatGPT that further evade detection to mimic human writing.
The argument of undetectable AI will probably be passé in due time, as it's not really going to matter in a few years. Content is content, and good content is what matters most, right? This goes for any field/discipline.
As AI becomes more integrated into our daily lives, we must continue to balance the power and potential of it to maintain integrity in the digital landscape. But I'm not quite sure how this will play out. Again, good content is really all that matters, so why would it matter how something was written if it answered everything that it needed to?
What are your thoughts? Have any questions? Drop us a comment below!
