Posted on Leave a comment

GPT-4 vs AI Detection

GPT-4

"GPT-4 can solve difficult problems with greater accuracy, thanks to its broader general knowledge and advanced reasoning capabilities." – OpenAI

GPT-4 Test

In this test I’ll be using Poe AI to access GPT-4 for free as currently at the time of writing it is a paid GPT model.

GPT-4 Initial Prompt

The initial prompt is as follows

Explain quantum computing in simple terms with high perplexity and burstiness in your tone of writing. Write at college level with an writing style of a professor.

GPT-4 Response

The response by GPT-4 is as follows

Ah, quantum computing! A veritable cornucopia of enigmatic conundrums and tantalizing paradoxes, nestled within the realm of the infinitesimally minuscule. Let us embark upon a journey, my dear intellectual companions, to explore the confounding depths of this wondrous domain.

Picture, if you will, the classical computer: a binary behemoth, churning out calculations with unyielding adherence to the 0s and 1s - the bits, as it were. A tale as old as time, or at least as old as the venerable Mr. Turing and his groundbreaking escapades. However, the quantum world scoffs at such simplistic binary constructs, choosing instead to embrace the enigmatic qubits.

Ah, the qubit! A mercurial protagonist in this grand narrative, dually embodying both 0 and 1 in a superpositioned waltz of mathematical probabilities. How does one even begin to fathom such a concept? Perhaps Schrödinger's mystified feline can offer us a clue, forever trapped betwixt life and death in its unobserved quantum dwelling.

Now, enter the quantum computing stage: a cacophony of entangled qubits, their fates inextricably intertwined like a cosmic dance of interstellar ballerinas. The sheer magnitude of computational prowess unleashed by these enigmatic entities is nothing short of breathtaking. The powers of parallelism are harnessed to perform unfathomable feats, solving problems in mere moments that classical computers could only dream of conquering in millennia.

But lo! The path to quantum supremacy is not without its pitfalls and perils. For the act of observing these capricious qubits invariably collapses their probabilistic waveforms, like a fragile soufflé deflating under the harsh gaze of an overly critical chef. The delicate art of quantum error correction, therefore, is of paramount importance, lest our dreams of quantum nirvana dissolve into a cacophonous cacophony of computational chaos.

In conclusion, my esteemed colleagues, quantum computing is an enigma wrapped in a paradox, nestled within the heart of the infinitesimally small. It stands as a testament to human ingenuity, challenging our very understanding of reality and pushing the boundaries of what is computationally possible.
GPT-4

GPT-4 AI Detection

The AI detection of choice for this test will be GPTZero as they claim to be "The World's #1 AI Detector".

GPTZero Result

Results

GPT-4 blows any AI Detection out of the waters. GPT models will continue to bypass AI detectors due to how hard it is to predict text being AI generated. If you want to get free access to GPT-4 consider downloading Poe AI which is the app I’ve used for this testing, you can also visit poe.com

Posted on Leave a comment

Transvribe – Educational AI for Youtube

Introduction

Transvribe is an AI-powered tool built by Zahid to help users learn more effectively on YouTube. With its powerful AI embeddings, Transvribe makes it easy for users to search for any video and quickly find the information they need.

Transvribe allows users to paste a YouTube URL and start asking questions. This makes it easier than ever to get the answers you need and learn more effectively. While cutting down on time for those long TED talks that you might have to take notes on.

Get Started with Transvribe

To get started with Transvribe, simply visit the website.

Transvribe.com

From there, you will need to paste the URL for the video, then you can begin asking questions, and taking notes right within the platform.

Posted on Leave a comment

ChatGPT for Google Slides

AI Generated Google Slides

MagicSlides App is an AI-powered Google Slides extension that allows users to generate full Google Slide presentations with just two key pieces of information. Just provide title, Number of Slides and some additional text which you want included (optional) and click generate and done!

Installation:

Go to installation page

Click on the Install button

MagicSlides App

Click on Continue

MagicSlides Requesting Permission

Select your Google account

Selecting An Google Account

Allow the requested permissions

MagicSlides Google Account Access

Done!

MagicSlides Installed
Posted on Leave a comment

GPTZero Case Study (Exploring False Positives)

Introduction

In this case study, I will be sharing the vast amounts of false positives current AI detection software gives, specifically for this case study I will be demonstrating GPTZero. I personally want to thank the supposed “Healthcare professional” who brought this to my attention via my contact link. It has motivated me to look more into this issue rather than just posting bypasses to these popular AI detection software programs, it will be only more beneficial to highlight their real usability in general.

What is GPTZero?

GPTZero is a tool that uses certain characteristics of AI to identify when AI is being used. It looks at things like how difficult it is for the AI to understand certain words or phrases, and uses this information to tell if the AI was involved in creating a piece of text. It’s like a “fingerprint” for AI, that can detect when it has been used.

The False Positives

From my own testing before even thinking about writing this case study on GPTZero, I have found that GPTZero gives very mixed results while OpenAI’s AI Classifier is usually more consistent but less informative about what is AI written and what is human written. After testing the medical paper on "Severe Outcomes Among Patients with Coronavirus Disease 2019 (COVID-19) — United States, February 12–March 16, 2020"Source

GPTZero states that more than 50% of this paper is AI written, and keep in mind this is only the first paragraph taken from the paper. Considering this is an official medical paper and that was written in 2020 by Multiple CDC Contributors. It’s not AI written, it’s quite obvious this is a false positive.

Note: Even after testing the maximum amount of words (5,000 characters) against GPTZero, the false positives still scored above 50%

GPTZero Result

Thank you to the supposed "Healthcare professional" for sharing this link

Pie Chart

The following pie chart follows these rules below.

  • All papers tested will be under the “neurology” topic
  • All papers that are flagged as AI or low perplexity counts will be marked as false positive
  • All papers abstracts will only be tested

The Results are shocking, as GPTZero marks 11 of the 20 medical papers as having AI written portions. Keep in mind, I only tested the abstracts of the papers.

GPTZero False Positives

Pie Chart Resources

False Positives:

Positives:

Wayback Machine Screenshot Link

This screenshot serves as a representation of what order these medical papers where tested in. The order follows from 1 through 20.

https://web.archive.org/web/20230219025214/http://web.archive.org/screenshot/https://www.ncbi.nlm.nih.gov/pmc/?term=neurology

The Control (Baseline)

As a control for our case study, I will be using OpenAI’s (ChatGPT Parent Company) AI Classifier.

Note: OpenAI's AI classifier is a "work-in-progress"

The image below shows the first paragraph from the medical paper on "Severe Outcomes Among Patients with Coronavirus Disease 2019 (COVID-19) — United States, February 12–March 16, 2020"Source

OpenAI’s Classifier Result

"The classifier considers the text to be unclear if it is AI-generated." - OpenAI's AI Classifier

Due to this paper being published in 2020 and the academic level of this paper. It’s extremely unlikely this paper used any AI based software. We can actually do another test to insure with 100% certainty, this paper didn’t use any AI when writing by looking back at it’s archived links via Wayback Machine by Internet Archive.

According to the Wayback Machine from the archived link https://web.archive.org/web/20210823160148/https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7725513/, dated (August 23, 2021). The wording has not changed since publication.

The Wayback Machine is a community based universal internet archive tool, that takes timestamped screenshots of the web. With the goal of perserving the web and it's content

OpenAI AI Classifier Limitations

“Current limitations:

  • Requires a minimum of 1,000 characters, which is approximately 150 – 250 words.
  • The classifier isn’t always accurate; it can mislabel both AI-generated and human-written text.
  • AI-generated text can be edited easily to evade the classifier.
  • The classifier is likely to get things wrong on text written by children and on text not in English, because it was primarily trained on English content written by adults.” – Source

The Problem

Having inaccurate reports on a software program that is about to go commercial is a big red flag and brings along many problems. Far too many problems to simply state in a single blog post. The biggest problem I personally can see it having and one that can relate to my life is plagiarism in the education field. If a student is wrongfully accused of plagiarism due to the institution following a flawed AI detection program, it could be detrimental to any student. At my institution, plagiarism carries a huge impact on your character. My university labels it as “Academic dishonesty, therefore, is a serious breach of university standards and will result in substantial penalties.”

Posted on Leave a comment

Bypassing ChatGPT Detection Programs Using Font Changers

What is GPTZero?

GPTZero is a tool that uses certain characteristics of AI to identify when AI is being used. It looks at things like how difficult it is for the AI to understand certain words or phrases, and uses this information to tell if the AI was involved in creating a piece of text. It’s like a “fingerprint” for AI, that can detect when it has been used.

Premium AI Detection Bypasses:

Introduction

Ever since GPTZero made headlines, I’ve tested the AI ChatGPT detection program to look for loopholes. I even created a Python script to bypass this detection tool. Funny enough, all that’s not needed, as all you need to bypass these popular AI detection software is an online font changer.

Note: Technically these aren't different fonts but rather they are different characters, this works because computers distinguish characters of different styles differently, but we'll call it fonts cause it brings better SEO

GPTZero Limitations

In my honest opinion, GPTZero is a very easy tool for people to bypass. However, in reality, people don’t even have to bypass it because it gives highly inaccurate results. It’s shocking to me that educational institutions trust this tool at all.

Step-By-Step Guide

Here’s a short, simple guide to bypassing many popular ChatGPT detection software programs using a free online font changer. For this guide, I’ll demonstrate it against GPTZero, as it’s the most popular tool among them all.

Generate your text with ChatGPT then copy your generated text, for my guide I’ll be using the text below

 Quantum computing is a type of computing that uses quantum-mechanical phenomena, such as superposition and entanglement, to perform operations on data. Unlike classical computers that use bits to represent information and perform calculations, quantum computers use quantum bits (qubits) which can be in multiple states at the same time. This allows quantum computers to perform certain calculations much faster than classical computers, making them well-suited for solving complex problems in fields like cryptography, chemistry, and machine learning.

Note: this string was generated by ChatGPT based on the question "Explain quantum computing in simple terms"

If we were to input a string of text generated by ChatGPT into GPTZero, it would likely classify the text as being generated by an AI.

AI Content Detected

Navigate to https://www.fontchanger.net/ with your copied AI generated text and paste it into the white text box

Font Changer Textbox

Note: If you are getting a lot of popup ad's consider installing uBlock Origin

Copy any of the fonts listed in Common Fonts

Font Changer Common Fonts

Lastly, test your newly font changed text into any AI detection software of your choice, for my guide I am using GPTZero

GPTZero Result

Video Guide

Posted on Leave a comment

Bypassing GPT-Zero’s ChatGPT Detection Program Using Different Font Types

What is GPT-Zero?

GPT-Zero is a tool that uses certain characteristics of AI to identify when AI is being used. It looks at things like how difficult it is for the AI to understand certain words or phrases, and uses this information to tell if the AI was involved in creating a piece of text. It’s like a “fingerprint” for AI, that can detect when it has been used.

Premium AI Detection Bypasses:

GPT-Zero: Limitations and Ways to Circumvent Them

Limitations:

GPT-Zero can identify responses generated by ChatGPT, however, its capabilities are limited. Currently, it cannot be used in real-world applications. If the need to bypass GPT-Zero arises, this is a simple method to do so.

Ways to Circumvent GPT-Zero Using Different Font Types:

You can use any text for my example, in this case I will be using a string generated by ChatGPT based on the question “Explain quantum computing in simple terms”.

Quantum computing is a type of computing that uses quantum-mechanical phenomena, such as superposition and entanglement, to perform operations on data. Unlike classical computers that use bits to represent information and perform calculations, quantum computers use quantum bits (qubits) which can be in multiple states at the same time. This allows quantum computers to perform certain calculations much faster than classical computers, making them well-suited for solving complex problems in fields like cryptography, chemistry, and machine learning.

If we were to input a string of text generated by ChatGPT into GPT-Zero, it would likely classify the text as being generated by an AI.

AI Content Detected

Note: You can test your text here. Be sure to select “Get Result” at the end of the page.

Step 1: Utilizing Python Script

Now that we know our freshly generated text from ChatGPT is detected by the algorithm of GPT-Zero, we can use a very simple tool to bypass GPT-Zero’s detection method. The tool we will use is a Python script that automatically replaces letters in sentences with the same letters but in a different font. It’s easy to use. Simply copy the script below

import random

def replace_letters(file_buffer):
    # Changes all AEOaeo letters with the cyrillic varients.
    file_buffer = file_buffer.replace('a', '𝚊')
    file_buffer = file_buffer.replace('q', '𝚚')
    file_buffer = file_buffer.replace('u', '𝚞')
    file_buffer = file_buffer.replace('n', '𝚗')
    file_buffer = file_buffer.replace('t', '𝚝')
    file_buffer = file_buffer.replace('m', '𝚖')
    file_buffer = file_buffer.replace('c', '𝚌')
    file_buffer = file_buffer.replace('o', '𝚘')
    file_buffer = file_buffer.replace('l', '𝚕 ')
    return file_buffer

def insert_zwj(file_buffer):
    output = ""
    zwj = "\u200D" # Zero Width Joiner unicode character
    for char in file_buffer:
        output += char
        # Randomly insert zero width joiner
        if random.random() < 0.2: # Lower number = Less ZWJ
            output += zwj
    return output


print(insert_zwj(replace_letters(input())))

input()

Note: To stay updated on the development of this script, visit my GitHub fork at https://github.com/gonzoknows/GPTZero-Bypasser

Step 2: Utilizing Online Python

Once you have copied the script above, simply paste it into https://www.online-python.com/ to utilize it online. Example image will be shown below

Online Python

To proceed, follow these steps in order:

Press the green run button

Online Python Run Button

Paste your text into the white input field

Online Python Result Field

Press enter when done

Press the red stop button

Online Python Red Stop Button

Press copy to clipboard button

Online Python Copy Button

Paste the output into a text editor or notepad

Copy the second output

Output Text File

Note: If you are getting a lot of popup ad's consider installing uBlock Origin

Once your article has been rewritten by the python script, copy and paste your text into GPT-Zero then press Enter to start the analyzation process. Scroll to the bottom of the page and select “Get Result” you will notice we have successfully bypassed GPT-Zero’s detection method!

AI Content Undetected
Posted on Leave a comment

ChatGPT vs. Turnitin – Does the Turnitin Plagiarism Checker Detect ChatGPT?

What Is ChatGPT?

“ChatGPT is a large language model trained by OpenAI that can generate human-like text based on a given prompt. It can be used for a variety of natural language processing tasks, such as language translation, text summarization, and question answering.” – This response was generated by ChatGPT

What Is Turnitin?

Turnitin is a web-based plagiarism detection service that helps educators and students identify potential plagiarism in written work. It compares submitted papers to a database of academic content, as well as content on the internet, in order to identify any matching or similar text. This tool is commonly used by educators to check student papers for plagiarism, as well as by students to check their own work for potential issues before submitting it.

Does the Turnitin Plagiarism Checker Detect ChatGPT?

Short Answer: No, here’s why.

ChatGPT is a language model that is based on a machine learning algorithm. It learns from every response it generates, constantly fine-tuning itself to produce more human-like text. Imagine a classroom of 100 college students who are all taught the same information. If you were to ask them all the same question, they would all give different answers. ChatGPT works similarly, as it is fed with millions of data points, allowing it to come up with unique responses. However, traditional plagiarism detection software such as Turnitin will not be able to detect plagiarism from ChatGPT, as it can change its responses in a way that is not easily traceable. Additionally, even newer programs specifically designed to detect ChatGPT plagiarism, like GPTZero, will not be able to keep up with the vast amount of data these models are trained on. Ultimately, the company with the most data will have an advantage in this field, as seen with companies like Google, Facebook, Apple, and Tesla.

Data/Results

I have conducted an experiment where I used ChatGPT to generate 20 essays, and then submitted them to Turnitin for plagiarism detection. The results showed that all the essays received a passing score, with very low similarity to other sources. This suggests that using ChatGPT to generate essays will be able to bypass plagiarism detection software such as Turnitin, as the model’s ability to produce unique responses makes it difficult to detect plagiarism.

Turnitin Results

Extra – Sam Altman Interview

StrictlyVC in conversation with Sam Altman, part two (OpenAI)

Posted on Leave a comment

Bypassing GPT-Zero’s ChatGPT Detection Program

What is GPT-Zero?

GPT-Zero is a tool that uses certain characteristics of AI to identify when AI is being used. It looks at things like how difficult it is for the AI to understand certain words or phrases, and uses this information to tell if the AI was involved in creating a piece of text. It’s like a “fingerprint” for AI, that can detect when it has been used.

Premium AI Detection Bypasses:

GPT-Zero: Limitations and Ways to Circumvent Them

Limitations:

GPT-Zero can identify responses generated by ChatGPT, however, its capabilities are limited. Currently, it cannot be used in real-world applications. If the need to bypass GPT-Zero arises, this is a simple method to do so.

Ways to Circumvent GPT-Zero:

You can use any text for my example, in this case I will be using a string generated by ChatGPT based on the question “Explain decentralized networks?”.

A decentralized network is a network that is not controlled by any single central authority, but instead is spread across multiple participants. In a decentralized network, each participant has a copy of the network's data, and each participant has the ability to contribute to the network. This allows for increased security and resiliency, since there is no single point of failure, and it makes it more difficult for any one participant to control or manipulate the network. Some examples of decentralized networks include peer-to-peer file sharing networks, blockchain networks, and distributed computing networks.

If we were to input a string of text generated by ChatGPT into GPT-Zero, it would likely classify the text as being generated by an AI.

GPTZero First Result

Note: You can test your text here. Be sure to select “Get Result” at the end of the page.

Now that we know our freshly generated text from ChatGPT is detected by the algorithm of GPT-Zero, we can use a very simple tool to bypass GPT-Zero’s detection method. The old-fashioned non-AI tool called Article Rewrite will do the trick. All you have to do is simply paste your text in the “Word Changer” section and press the button “Rewrite Article”.

Article Rewriter

Note: If you are getting a lot of popup ad's consider installing uBlock Origin

Once your article has been rewritten by Article Rewrite, copy and paste your text into GPT-Zero then hold Ctrl and press Enter to start the analyzation process. Scroll to the bottom of the page and select “Get Result” you will notice we have successfully bypassed GPT-Zero’s detection method!

GPTZero Second Result

Posted on Leave a comment

How to Summarize YouTube Video Transcripts with ChatGPT

Summarizing long YouTube videos can be a tedious task, but it’s often necessary for quickly getting the main points of a video or for creating notes. One way to streamline the process is to use ChatGPT, a state-of-the-art language model developed by OpenAI, to generate a summary of the video in the form of bullet points.

Here’s a step-by-step guide on how to use ChatGPT to summarize any YouTube video with a transcript:

First, obtain the transcript of the YouTube video. You can usually find a transcript by clicking on the “Three Dots Icon” button next to the share & clip buttons. If the transcript is not available, you can use a transcription service to create one.

Show Transcript Button

Next, open ChatGPT in your web browser. You can access ChatGPT at this link: https://chat.openai.com/auth/login. Note: Currently, obtaining a chatGPT invite requires being on a waitlist due to the large number of users.

Once you have ChatGPT open, enter the following command to start the summarization process: “summarize transcript into bullet points.”

ChatGPT will prompt you to enter the transcript of the YouTube video. Copy and paste the transcript into the input field and hit “Enter.”

Timestamps in ChatGPT

ChatGPT will process the transcript and generate a summary in the form of bullet points. The summary will be displayed in the output field of ChatGPT.

ChatGPT Result 1

Review the summary and make any necessary edits. You can also adjust the length of the summary by using the “summarize transcript to X sentences” command, where X is the desired number of sentences in the summary.

ChatGPT Result 2

Once you are satisfied with the summary, you can copy and paste it into a document or save it for later use.

Using ChatGPT to summarize YouTube videos with transcripts can save you a lot of time and effort, and it’s a great way to quickly get a sense of the main points of a video.