Skip to main content

Week 3 - MBA 6101 - The Battle Between Educators and Artificial Intelligence


With the rise of Artificial Intelligence (AI) and Generative Pretrained Transformer (CPT) text, many teachers are faced with a new challenge. How do we prevent students from using these tools to complete their course work for them? The answer may be to use a tool developed to identify text that has been generated by these GPT tools. Open AI responded to concerns by creating their AI Text Classifier. The tool is very simple to use, simply copy and paste the text into an open text box and click a button. With options that range from very unlikely to likely created by AI, the user can quickly identify content that was generated with GPT.

However, one might ask how accurate are these tools really? I tested out several blog posts from this very blog to see how well the AI Text Classifier would perform at identifying those that I leveraged Open AI's ChatGPT to create. Interestingly, none of my blog posts were identified as having been likely generated using GPT. In fact, of the three posts that leveraged Open AI's ChatGPT, AI Text Classifier two as possibly written with GPT, and was unclear on one of the posts.


Perhaps even more interestingly, when I ran the blog posts through the same tool one post was identified as being very unlikely as having been generated with GPT while the second was also unclear as to whether it was generated using AI. The post listed as being unlikely written with AI was written in a first person format with general instructions on how to create a website using Google Sites. The post where AI Text Classifier was unclear on the source information was written in a more professional voice regarding how guerilla and traditional marketing strategies differ. I would argue that the post that was unclear on whether GPT was leveraged would be written in a similar manner to how students would be expected to write their assignments.



So, are there better tools available to identify GPT generated texts? One suggested option is GPTZero which indicated that parts of both of my completely human written blog posts as potentially containing AI generated text. In addition to the simple statement that the text potentially included AI generated text, it highlighted the specific sentences that triggered this assessment. It did correctly identify the GPT generated blog posts as having likely been entirely generated by GPT. This tool appears to be more on the conservative side of identifying GPT text, but the site does indicate that this should only be one tool for educators and that they should consider the students individually rather than relying solely on their results.



The best tool, in my opinion, is ZeroGPT which not only provides it's analysis as a simple statement but also provides a percentage of the text it believes could be AI/GPT generated. The three posts where I leveraged Open AI's Chat GPT were all correctly identified as being 100% AI/GPT generated. Although, there was editing done to each of these posts.



The manually written blog posts, while not listed as have a 0% chance of containing AI/GPT Generated text, provided the below images as well as highlighted lines to identify where the potentially generated text was in the document. And you could guess from the below images, the first one was the one that also was indicated as being unclear from Open AI's ChatGPT Text Classifier and potentially containing AI generated text by GPTZero. I find the visual provided by ZeroGPT with a possible percentage shows that this is not fool proof tool, it even provides a visual gradient to illustrate this fact.





Overall, tools exist to help educators identify potential plagiarism with GPT. But these educators should be careful when using them as they may over or under identify this potential. Ultimately, educators should be holistic in their evaluation of students which can be a complex process. 

Comments

Post a Comment

Popular posts from this blog

Week 2 - MBA 6101 - Creating a Website and Domain with Google Sites: As Easy as 123

As a beginner in website design, I would have thought it necessary to have some coding or design background to create my own website. Of course, I remember the days of Myspace where we all had our own "websites," using the most basic of HTLM codes to customize wallpaper or music. But Google Sites makes it even easier than those days of the past, with templates, simple blocking, options to upload images directly using a Google Image search, and even embedding links, all with a simple click of a button or two. You should know, a Google account is required to create a site using Google Sites. If you need assistance with setting up a Google account you can find information on that here . Once you have an account set up you will need to access your Google Drive . From here, you can create a multitude of documents, forms, or even a site. To create a website, simply click the "New" button, hover over "More," and click on "Google Sites."         From her
1 comment
Read more

Week 4 - MBA 6101 - A Marketing Plan: The Cornerstone of Your Success

According to Guerilla Marketing  by Jay Conrad Levinson, a good marketing plan can be the difference between success and failure for a business. He suggests that an entrepreneur put together their market plan before ever selling their first product or providing service for the first time. A marketing plan should be simple and straightforward, it will help define the identity of a business and how the company will interact with customers. The first step is to answer the below seven questions to develop a marketing plan to guide your business strategy. What is the purpose of marketing your product or service? Why should customers choose your product or services? Who is your target audience? What marketing tools will you use to meet your purpose? Where do you fit in within the market? How do you want customers to see or identify your company? How much are you willing to invest in your marketing plan? For instance, if you were looking to start consulting for a pharmaceutical company the be
Post a Comment
Read more

Week 2 - MBA 6101 - Artificial Intelligence and Machine Learning Application in Safety Oversight for Clinical Trials

Clinical trials are crucial for the development and approval of new medical treatments and drugs. However, ensuring the safety of the participants is of utmost importance. This is where the use of Artificial Intelligence (AI) and Machine Learning (ML) can play a significant role. These technologies can help streamline the safety oversight process and make it more effective, thereby improving the overall quality of clinical trials. Here are some of the key applications of AI and ML in safety oversight for clinical trials: 1.    Adverse Event Monitoring: AI algorithms can be trained to detect and identify adverse events that occur during a clinical trial. The algorithms can analyze large amounts of data quickly and accurately, which can help identify potential safety risks earlier and prevent adverse events from reoccurring. 2.   Predictive Modeling: AI algorithms can be used to develop predictive models that can help identify participants who are at higher risk of experiencing adverse e
Post a Comment
Read more