In the video: the artificial intelligence Chatgpt answers the request I wrote to it (screenshot)
The use of artificial intelligence in education is growing, and with it the fear of the ability to cheat in tests and assignments using this type of technology.
ChatGPT, a natural language processing (NLP) system, is an example of the use of artificial intelligence designed to promote the ways of learning and access to information sources.
However, its and other artificial intelligence systems' ability to provide answers during a conversation raises concerns about the possibility that it will be used to cheat.
This opening paragraph was not written by a human, but by the software itself.
In recent weeks it is the hottest topic in the academic world.
In universities around the world, emergency discussions have come together that deal with the consequences of launching ChatGPT to the general public.
"The WhatsApp group of our faculty members is full of discussions on the subject. It is very clear that this is a technology that changes the rules of the game," says Prof. Eran Toch, head of the undergraduate program for industrial engineering and management at Tel Aviv University.
The software, which is built as an internet chat, is currently open for free use.
You only need to register on the site, mark a V next to the sentence "I am not a robot" - and you are in.
You can try to make small talk with her, but her main potential is to be an extremely effective work tool, one that can help many professions.
Almost every question addressed to her will receive an answer: from educational topics such as how processes work in the human body and the stages of the First World War, to fateful questions such as what are the most beautiful cities in the world, and can men and women only be friends.
You can also ask her to write a rap song or a science fiction story, or use her to identify problems in the code.
One of the examples given by the software for a question that will present its capabilities is "Do you have any creative ideas for a tenth birthday?".
"Changes the rules of the game".
ChatGPT's software interface (photo: screenshot, OpenAI)
Behind the chatbot is the AI company, whose founders include Elon Musk.
The company whose name deals with artificial intelligence has also recently gained publicity through the DALLE-E software that creates visual images from text.
To bring the chatbot to its current level, the company put it through lengthy "training sessions" that included two types of learning.
The first was reinforcement learning, in which the practitioners rewarded the software for good answers through a high rating, and taught it to prioritize answers that would receive a high rating.
The second was guided learning - solving problems based on previous solutions.
When they felt it was ready, they released it to the market.
Not a week passed until more than a million users joined it.
The students also jumped at the bargain.
The notes in the pencil case, the summaries in the bathroom and the transfer of work from generation to generation instantly became a thing of the past - now it's easiest to cheat with a chatbot.
"I found myself entering the software during lectures and looking for the material there. She summarizes better than the lecturers," said a computer science student.
"I know a student who didn't know a certain code language and didn't have the strength to learn it, so he just wrote in the language he knew and asked her to translate it into the required language. One of its effective uses is to identify problems in the code you wrote."
"We're starting to use it ourselves," Toch admits.
"Professors write opinions on other studies and it takes up a huge part of our time. Usually the opinion begins with a description of what the article is about. ChatJipity does this part not bad at all. If we use it then surely students will too."
More in Walla!
The device that makes a revolution in the fight against wrinkles - in an introductory sale
Served on behalf of B Cure Laser
Prof. Eran Toch, Tel Aviv University (photo: official website, Shahar Shahar, Tel Aviv University spokeswoman)
The secret does not remain within the walls of higher education institutions.
Since the launch of the TikTok software as a concoction.
Videos about videos that deal with chat bots are uploaded to the social network and get tons of likes and comments.
Some of them teach how to use her in the most efficient way, show how she summarizes the material and which questions she answers best.
Others simply feature her answers to funny questions or pearls of wisdom like five sentences about what it's like to get rich.
The youth hear the advice and try to implement it themselves.
The effects can be far-reaching.
The teachers and lecturers got used to being satisfied with generic, self-explanatory questions, often ones that appeared in previous works.
Now, pupils and students will be able to return such work with the effort of a few clicks on the phone and get a satisfactory grade.
The global education system will have to exercise its creativity - and demand the same from the other side.
"Nowadays, all learning, both in elementary, high school and university, goes in the direction of standardization. But artificial intelligence knows how to think exactly like that and sets a limit to standardization," says Toch.
This will also have budgetary implications.
"The financial expenditure will increase. It will not be possible to hire a practitioner to check everything, or to take a multiple-choice questionnaire, put it in Excel and have it checked automatically. The classic methods of academic papers are becoming less efficient. The idea of a student who writes a paper, polishes paragraphs and gets a grade is going to change."
ChatGPT's software interface (Photo: ShutterStock)
So are the days of frantic searches and crossings between books, summaries and Google over?
The answer is probably not yet.
On the face of it, the text provided by the software seems logical.
The sentences are correct, the connection between the paragraphs is coherent and spelling errors will never happen.
It is difficult to distinguish it from the formulation of a human being, if he himself knows how to write.
The accuracy of the content is another story.
As long as the information it must provide is on the Internet in an orderly manner, or is as technical as writing code, the result will be satisfactory.
But a more complex task will sometimes lead to illogical texts that only feel like a correct answer.
Another and more intractable issue is that it is based on events that happened up to 2021, so challenging it with current news won't work.
"There is a fundamental problem with the output that is not going to be solved," clarifies Prof. Toch.
"The algorithm summarizes the Internet and instead of search results creates a new text. This means that if it is not on the Internet it does not exist, it does not invent things by itself. Sometimes the texts are also strange. They can include specific numbers because they were written on one page on the Internet. Or they are terribly general and uninteresting , or too specific and require a lot of editing. Until she can answer questions reliably, she will have to go through several more incarnations."
The computer science student agrees with his words.
"I did an experiment: I sent Chat a five-page assignment to see how he answers. I discovered that he cannot be trusted 'blindly'. "It's still like Google Translate - does 80% of the job right and 20% nonsense that needs to be corrected."
news
News in Israel
Education
Tags
artificial intelligence