Week 1 — Success of AI Writers
AI — Artificial Intelligence
As a start of a weekly project blog post series, I want to briefly mention my teammate Hüseyin Yiğit Ülker and myself. We are right now in our junior year at Hacettepe University Artificial Intelligence Engineering. We are doing an artificial intelligence project in the AIN311 course, which is one of the most important courses in our department. From this week until the end of the project, we will share with you all our progress as a blog post. In addition to keeping a record of our progress with these blog posts, we also aim to contribute to the computer science society. At the same time, thanks to the feedback from you who follow this series, we will have the chance to see our mistakes and progress faster.
AI writers are today one of the most popular topics. Considering their technologies and success rate, they are very charming to AI nerds. With AI writers, you don’t have to hire a writer and look for cheap hacks; the tool does all the work for you! Not only does it cater to all your writing needs, but it also provides you with powerful analytics that helps you predict and optimize your content. This unique feature saves you hours in research, which is a bonus when you consider how much time and effort it takes to find good writers.
Amazing right? I had the AI writer write the opening paragraph to show you what this incredible tool can do. By writing just one line of text. You can find more content written by AI writers and their analyses from different angles in the news below.
In recent years, artificial intelligence (AI) has made incredible strides in its ability to generate human-like text. As a result, AI writing is becoming increasingly commonplace, with businesses and organisations…
These tools use OpenAI’s deep-learning AI writing model, Generative Pre-trained Transformer 3 (GPT-3). Creates new sentences and paragraphs based on a few statements you give. Moreover, the GPT-3 model is also very successful in many other tasks:
- Rephrasing content
- Spell checking
- Grammar checking
- Researching topics
- Generating ideas
- Completing forms
- Preparing reports
- Formatting documents
Honestly, it is a very successful tool for students who have to write in many different types(reports, assignments, articles, etc.). As a student, if you want to go effortlessly, it will do it quite successfully for you. Can get high grades for you with minor corrections. The example in the news is one of the most realistic expressions of this.
The student essay written by an AI writer about “Learning Styles” in the blog post published by Michael Sharples, Emeritus Professor at Open University in the United Kingdom, and the Emeritus Professor’s thoughts on this subject resulted in a completely different problem coming to our minds:
PLAGIARISM
Plagiarism is taking the words or ideas of others and using them as originals and presenting them without permission.[1]
There is an increase in plagiarism cases with the development of technology. During the pandemic period, many countries switched to online education and the use of technological devices (computers, tablets, phones) increased a lot during this period. According to the research conducted by Copyleaks in this period, the rate of plagiarism increased by 10% from 35% to 45%.
Therefore, there are various plagiarism detection programs available to prevent plagiarism.
Turnitin, Quetext, Grammarly…
Since AI writers shine when paraphrasing sentences, we start thinking that these tools can be used by people with bad intentions for plagiarism. A scammer can take the work that you have written with a lot of effort to paraphrase with a few keystrokes and publish it as if it were its work. Since we could not find a study on how existing plagiarism applications work against AI writers, we decided to research this subject ourselves.
Let’s talk briefly about what we plan to do without mentioning too many technical details. Firstly, we are planning to develop several different artificial intelligence models for plagiarism detection using methods already in the literature. For this step, we are going to use a publicly available plagiarism dataset for training and testing these models. Since there is no data obtained with AI writers in the present datasets, we will manually paraphrase selected articles with AI writers to observe the process, which is our main concern. Then, we will test the existing models and examine the results with the duplicate data we have obtained. For future work, if we can paraphrase enough articles, we can retrain and test existing models using transfer learning. This step is not among our current goals due to both the schedule and the difficulties in generating data.
Our main purpose in this study is to measure the success of AI writers in plagiarism detection applications.
Wait until next week for more technical details. After all, this will be a technical series. This week we wanted to talk about the problem and terms. Also, this is the first blog post we’ve published. We get it, it’s very difficult to even write that much… And there was no energy leftover to talk about technical details :) It is so hard to write something, to put thoughts into writing. But as can be understood from our mistakes, we did not use AI writers in this blog.
Stay Tuned
