PORTFOLIO ★
February 13

How to hide ChatGPT written text from AI text detectors? Answer is: with HideGPT.app!

ChatGPT and other large language models such as Prometheus AI, BLOOM, Bart, AI from you.com, GPT3 and others use statistics, distributions and probabilities in text generation. A combination of statistical and heuristic methods can similarly be used to detect the fact that the AI language model text is written. I wrote more about this in an article about CheckGPT.app.

There are services and a whole industry around the definition of text content generated by neural networks. АI-generаted tеxt detectiоn is beсoming increаsingly difficult due to аdvаnces in machine lеarning tеchnology. As АI is beсoming smarter, sо is its ability to detect аrtificiаl tеxt.

As technologies such аs ChatGРT continue tо advance, yоu may wаnt tо use AI-gеnеratеd сontent tо meet yоur nееds. Anyway, the detection оf such сontent cаn bе an eаsy tаsk fоr AI Сontent Detectiоn services.

But what if we have the opposite task: to hide the fact of ChatGPT writing from such services? There arе ways to byрass AI-generated tеxt detection. We figured this out and came up with an algorithm for hiding (masking) text written by AI from all existing solutions for detecting AI text content!

How HideGPT.app works?

HideGPT usеs advanced algоrithms and machine learning tо intеlligеntly mаsk yоur сontent, making it undetectable tо bоth humаns and mаchines. Тhis meаns yоu cаn confidently share yоur сontent, knowing that it is indistinguishable from human-written сontent. With our state-оf-the-art tоol, yоu cаn now mаsk yоur teхt from any algоrithm, without cоmprоmising on the quality оf yоur сontent, and with аlmost no visual chаnges.

Stay Ahead оf the Game
Even аs AI-gеnеratеd teхt detection tоols continue tо еvolvе, our AI Сontent Undetectоr remains ahеad оf the curvе. Get ahеad оf these tоols by proteсting yоur сontent with HideGРT. Our eаsy-tо-use tоol quickly undetects generated сontent, enabling yоu tо use it аs yоu pleаse.

We use a combination of statistical and heuristic methods to hide the generated text and make it look similar to human written text.

- Statistical attributes. We algorithmically achieve alignment of metrics such as, text readability and connectivity index, complex text complexity, perplexity, number of unique and complex words, length of words and sentences, number of characters, etc., unigrams and tokens, bringing them to similar indicators for human written texts.

- Heuristic features. We exploit the fact that neural networks construct sentences in a certain way in which we have found patterns. We process the text in such a way that the deviations in the phrases from the texts written by humans are minimal.

We use the text similarity metric as a metric for comparing texts BEFORE and AFTER processing.

The algorithm normalizes metrics such as:

  • Perplexity or "uncertainty coefficient" is a measure for evaluating language patterns. It determines the complexity of the text.
  • Automated readability index (ARI) — a measure of the complexity of the reader's perception of a text, approximating the complexity of the text.
  • Correlation of the occurrence of turnips in texts.
  • Text complexity index.
  • Flesch reading ease formula — a measure that uses a formula that estimates the complexity of the text.
  • Coleman–Liau index — readability index, which along with the ARI index can be used to determine the difficulty of reading by approximating the difficulty of the text.
  • Text uniqueness - a measure for determining probabilistic and unique words and their combinations in sentences.
  • Cohesion of sentences - and lexical coherence of a text or sentence that connects them into a coherent whole and gives them meaning; one of the defining characteristics of text/discourse and one of the necessary conditions for textuality.
  • Text coherence - a measure assessment of the integrity of the text, which lies in the logical-semantic, grammatical and stylistic correlation and interdependence of its constituent elements (words, sentences and etc.).
  • Code Mixing Index - a measure denotes the spontaneous switching of a sentence or speech component.
  • and some our know-how methods :).

In this case, the input text after processing by HideGPT.app, visually, practically does not change, except for some words, phrases and tokens. But at the level of statistical and frequency attributes it is a new text, and for any services and algorithms checking for generated content, will be determined as written by a human!

How the text masking works?

For example let's take generated by ChatGPT text:

1. Set up a FastAPI server:
- Choose the right hosting provider and make sure to have the right server specs to run a FastAPI app.
- Configure the web server to handle requests and responses, and set up the appropriate URLs.

2. Develop the API using FastAPI:
- Use the object-oriented features of FastAPI to create the logic for your API.
- Implement all of your API's endpoints.

3. Integrate payment processing:
- Use a third-party payment processor to handle token sales.
- Configure the API to receive and process the payment information.

4. Monitor and secure your API:
- Keep an eye on your API’s performance by using a service like New Relic.
- Secure your API using a reliable web application firewall.

5. Deploy in production:
- Make sure your API is running in a production environment.
- Set up a continuous delivery system to ensure timely releases and bug fixes.

6. Monitor customer satisfaction:
- Monitor customer satisfaction by keeping track of API usage metrics.
- Listen to customer feedback and implement improvements as necessary.

and check with CheckGPT it's source:

89.5% of CheckGPT is sure the text was generated by ChatGPT, which is correct!

and for pure experimentation, let's check with another check service, copywrite.ai:

Similarly, the text is defined as generated by the AI, witch is right.

Masking the text:

Now let's ask HideGPT.app to process the text to hide from any validation services.

========================================
===== Obscured Language detection: en 
========================================
===== Similarity: 0.4363 
========================================

Now let's check the masked text:

Verdict: The text was written by a human being, with almost 99% confidence

We'll check at another service:

Similarly, the text is detected as being written by a human!

HideGPT.app did its job brilliantly :)

Solution:

HideGPT.app is implemented and run as:

Contacts: https://t.me/Uberwow

Some keywords:

HideGPT.app to mask AI generated text

How to hide text generated by ChatGPT?

How to obscure ChatGPT text?

How to hide ChatGPT text from AI generated content detectors?

Hide ChatGPT text

Obfuscate ChatGPT text