ChatGPT ruins business for college job pushers

As every year, the arrival of spring floods the Internet with announcements of companies that sell university theses, especially TFG (Final Degree Project) and TFM (Final Master’s Project). They trade freely. “There are two types of people, those who are living their lives and those who are doing the TFG,” says one of them, posted on Instagram. “Possibility of payment in installments. Don’t let the price be your excuse!” promotes another on Google. “Do the internships consume you every hour of the day?” asks the umpteenth person who appears in the Facebook ad library: “We can help you.”



A fake Drake song made with artificial intelligence triggers alarms in the music industry

Further

These companies are defined as “academies” or specialized “in content and academic research” and are legally constituted. Many have several employees, who divide the work between those who write the projects and the intermediaries who serve as liaisons with the students. They make it easy: they have company WhatsApp accounts with which their “academic advisors” ask for the necessary details of the project and close the deal with the students.

Dissertation trafficking operates in limbo, since it is not illegal for these firms to sell a TFG or TFM. That the student then submit it in his name is another matter, but there the universities have little more to do than review it with anti-plagiarism programs. If the dreaded Turnitin (the most powerful and well-known tool in this category) approves the text, everyone is happy. elDiario.es has asked the Ministry of Universities if it has any protocol in this regard, but has not received a response.



But not everything is rosy when it comes to retailing final degree projects. These types of services are not cheap. The bill for a black market TFG or TFM depends to a large extent on the degree, the extent of the work and what the student can contribute from their own harvest to serve as a basis for black market writers. But they move in several hundred euros, at least.

Many of these companies also accumulate negative reviews. “I paid 440 euros as soon as they asked me, before carrying out the work and they have made me waste my time and my money. The wording of the disastrous work, the useless content, the bibliographical references, few and old”, protests a dissatisfied customer on the review page of one of these companies on Google. “A complete scam. I have spent €860 on a copy and paste having an 85% similarity with another colleague, in addition to not making the corrections, not putting regulations or laws or anything, the writings are inconsistent and if someone wants the tests I have them to prove it ” laments another.



In this course, however, a new actor has appeared in the equation. “In my class almost everyone is looking to see if they can do some part of the TFG with ChatGPT,” admits Manuel (not his real name), a student of Law, Labor Relations and Human Resources at the Complutense University. OpenAI’s text-generative artificial intelligence already caused a schism in education immediately after its release in November 2022. This end of course will be the first where ChatGPT will be fully available to students and could help with writing of thousands of final projects.

Is ChatGPT able to complete in minutes a TFG designed to involve a job of several weeks or months for the student just by asking? No. You have to know how to relate to the tool to exploit its writing capabilities and that they fit with this type of academic work. But the clues on how to do it are already running like wildfire through social networks. This video of a young woman published in April on TikTok has already accumulated one and a half million views.

To this we must add that ChatGPT is still free, although OpenAI limited it to 10 messages per day for people who do not pay a subscription. That fee, however, is 24 dollars a month (about 22 euros). Much cheaper than what the TFG and TFM writing companies charge, which can be paid among several colleagues and which also gives access to GPT-4, a more advanced model than the original ChatGPT.

No countermeasures to ChatGPT

Faced with the controversy caused by its possible negative impact on education, OpenAI launched a tool to detect texts written with artificial intelligence. However, this only works in English and it is rare that it offers conclusive results, as this medium was able to verify after asking it to analyze various texts. Their conclusions, moreover, are never definitive and range from “very unlikely”, “improbable” or “confusing” to “likely” or “very likely” to assess the intervention of an AI in a text.

“The classifier is likely to err on texts written by children and on texts that are not in English, since it was mainly trained on English content written by adults,” OpenAI acknowledges. This will be the big problem of the programs that detect plagiarism or artificial intelligence writing at the end of the 2023 course: they are not prepared to be blunt when the text to be analyzed is in Spanish.

Manuel explains that some of his Complutense professors have warned them that tools like Turnitin are capable of catching cheaters both in plagiarism and with AI. The truth is that the company accelerated to incorporate this function into its detector before this end of the course and achieved it on April 4. But not in Spanish. “At the moment this tool is only available for text in English,” company sources explain to elDiario.es.

To this is added that locating text written by a machine is not as definitive as plagiarism or paraphrasing, the company assures: “Since our false positive rate is not zero, you, as an instructor, will have to apply your professional judgment, the knowledge you have of your students and the specific context of the task”.

The Ministry of Universities and the Conference of University Rectors have not moved either to provide a uniform protocol to the centers. Both institutions met in February to discuss the impact of ChatGPT on academic work and agreed to follow the instructions of the European Association of Universities, which recommends the creation of working groups to analyze “the immediate consequences of Artificial Intelligence tools in teaching and learning processes, with particular urgency in relation to evaluation procedures”.

Sources from the Ministry headed by Joan Subirats maintain that, for the moment, there have been no updates on this matter.

“To see them come”

Faced with this situation, many teachers have stayed “a bit to see them coming” in relation to the use of AI in the TFG and TFM this 2023, acknowledges María del Mar Sánchez, a researcher specialized in Educational Technology and professor of Didactics and School Organization. from the University of Murcia: “I had the exposition of the TFG yesterday [por el jueves] and it is true that internally, among colleagues, there was concern about ChatGPT”.

However, the teacher calls to relativize the situation and not be alarmist. The traps are not new. “There have always been companies that can be used to do this work and some students have copied from work from other years,” she recalls. “On the other hand, we tend to make it look like students are evil and they ultimately make use of the resources at their disposal in the most efficient way, as we all would,” she says, referring to AI.

We tend to make students see that they are evil and they, in the end, make use of the resources at their disposal.

Sánchez knows that some of the students who have presented their theses this week have used ChatGPT to try to save work. “You can see it in his defense of the job,” he reveals. Other students have confessed that they have used it as a support, to give them ideas or improve the clarity of their writing. “I think that’s not bad either. There are students who like to investigate and artificial intelligence can be just one more tool, ”he says.

Lastly, during her conversation with this medium, the teacher highlighted that several of the conversations about renewing the education model had surfaced long before the advent of artificial intelligence. “What ChatGPT has shown is that there is no alternative,” he stresses: “What we have to do is give the research model a very deep turn, establish more tutorials to be able to talk with them and verify that they are obviously doing the process of defense investigation. Producing TFG and TFM wholesale is not going to make much sense”.

Related articles