How ChatGPT Might Give Teachers Headaches


Mathieu Grumiaux

January 03, 2023 at 08:00

41

artificial_intelligence1600

the chatbot ChatGPT could offer results so effective that they would trick the anti-cheating systems implemented by teachers.

ChatGPT’s prowess has been widely commented on in recent weeks and the tool is so powerful that it has the education world intriguing.

An almost perfect cheat tool for students

If the OpenAI chatbot is not perfect and can sometimes give absurd answers, the tool is now powerful enough to provide detailed explanations and reasoning on a large number of subjects, even the most advanced. A feat that seems to scare Google, which is worried about the lead taken by its competitor.

Pupils and students of the world should not be mistaken and use it very quickly to do their homework for them, without the need to delve into books, or to think about the meaning of their essay.

Professors take this problem very seriously and seem at the moment helpless in the face of the constant progress made by OpenAI in its technology. If the latter had, over the years, found methods to quickly detect copy-paste from Wikipedia, ChatGPT goes much further, with a more in-depth analysis offered for each question.

OpenAI and faculty are already considering measures to better spot assignments written with ChatGPT

What caught me off guard was how big a leap it was said Eric Wang, vice president of artificial intelligence for Turnitin, a startup founded in 1999 and working with 16,000 school systems around the world to build and update anti-cheat software of the same name. .

At the moment ChatGPT seems limited, and its limited turns of phrase or vocabulary are recognizable when compared to a typical student’s prose. However, the technology continues to advance and could be much more difficult to detect in just one or two years.

OpenAI is aware of the problem and is looking for solutions to help teachers track down plagiarism, without limiting the development of their artificial intelligence. Scott Aaronson, professor of computer science at the University of Texas, Austin, and visiting scholar at OpenAI, reportedly said that the company is working on introducing cryptographic signals that are invisible, except for specialized companies like Turnitin.

Teachers could also be called upon, by modifying the title of the subjects, to take account of recent news, an area which is not yet supported by ChatGPT, or by asking students to justify their reasoning by video or in front of the teacher.

Source : Bloomberg



Source link -99