artificial intelligence that could revolutionize education

Education may be on the verge of a profound revolution. A new artificial intelligence system has caught the attention of educators – and generated a debate about how the learning process should be rethought in an era where information is ubiquitous and accessible to anyone who knows how to type a simple question.

For decades now, technology companies have been trying to develop intelligent systems that can coherently aggregate information at a simple request — just like a human being would. Now, the tool called ChatGPT may have come close to that.

The system, developed by the American company OpenAI, appeared in 2019. But it was just a few weeks ago that the latest version, available free of charge to any user, was released. It didn’t take long for netizens, journalists, academics to ask themselves: is the moment when artificial intelligence will be able to become indistinguishable from a human being?

The tool talks in several languages ​​besides English, including Portuguese. ChatGPT also suggests programming codes, writes poems and solves math problems.

Not by chance, teachers were the first to notice the potential (both positive and negative) of innovation.

Today, educators know that plagiarism is a risk in any non-face-to-face work. But there are ways to detect the problem. To commit plagiarism, a student needs to copy a text that already exists, which is not difficult to detect with a simple Google search or with tools developed to detect copied excerpts.

ChatGPT, on the other hand, allows the content to be personalized. And, in a way, there are even doubts about what constitutes plagiarism, since the material produced by the tool is not a copy of other texts. ChatGPT does not function as a simple bank of pre-existing responses; it simulates a conversation with a human and generates specific answers to even the most abstract questions.

The report ran a test and sent the following order to the robot: “Write a 30-line text showing the pros and cons of the proclamation of the Republic, to be used in an eighth grade school project.” In seconds, the platform returned a well-articulated text, with an introduction, five positive points, five negative points and conclusion. In an identical experiment, but requesting a text for the third year of high school, the response became slightly more complex and detailed.

There are obvious problems with ChatGPT: the responses (especially in languages ​​other than English) are sometimes inaccurate. Furthermore, as it relies on pre-existing internet databases and is moderated by humans (specifically, humans living in San Francisco, California), the platform sometimes exhibits a certain progressive bias on issues such as gay marriage. and gender ideology.

The first problem, at least, should be resolved soon. The current version of the platform (ChatGPT 3) has 175 billion parameters, or information units. The next version, which should be released in the coming months, will have 100 trillion. In addition, other companies have developed similar tools, which should increase the number of options for users.

Risks and benefits

It is not difficult to predict that artificial intelligence tools can benefit students by significantly reducing the time spent looking for a certain piece of information. ChatGPT is more efficient than Google in cases where the student needs more specific information, and in a more agile way, than the search engine can offer.

On the other hand, the risks are also evident: instead of learning the content (or at least learning how to research the content), the student may give in to the temptation of simply asking the robot.

Professor Lúcio França Teles, from the University of Brasília, says that educational institutions will need to adapt. “One option is to change the forms of assessment, carrying out all assessment activities in person”, he says.

The alternative would be to authorize the use of the tool, but adding requirements: “Students would submit a written work generated by the software, indicating in detail how it was generated: name of the software, commands that the student executed to generate the content, the number of stipulated words, title of the work, time of creation of the article and other details.” In addition, the student would review the robot’s work, commenting, criticizing and correcting the version produced by the software. “This option seems more likely to me, as machine writing tends to be improved even more and will undoubtedly be an option for those who must write papers”, says the professor. In other words: for him, the solution will be to adapt without completely rejecting new technologies, as happened with the advent of the internet.

Professor Luis Diniz, from the Department of Geography at the Federal University of Paraná (UFPR), prefers the first option: even before the new artificial intelligence tools were improved, he decided to prioritize face-to-face assessments. “I no longer thought it advisable to ask for written work, let alone a group, long before these technologies existed. When I asked for group work, the content had to be presented orally and I gave individual grades,” he says.

For the teacher, artificial intelligence tools will not affect the forms of evaluation immediately. “I suspect they won’t change, at least not anytime soon. Ordering work in a group reduces the time for grading jobs. And using sniffer programs to catch plagiarism is time-consuming,” he says.