ChatGPT Releases New Tool to Identify AI Writing

Reading audio



04 February 2023

A new tool from the maker of ChatGPT aims to help deal with concerns about how artificial intelligence (or AI) can be used to cheat in school.

The tool is called AI Text Classifier. It is designed to identify writing that was produced not by students but by AI programs.

The tool was launched by OpenAI, an AI technology company based in San Francisco. The company is the maker of ChatGPT, an AI system that can produce any kind of writing on demand. Many education officials are concerned that ChatGPT could fuel academic dishonesty and harm learning.

But, OpenAI warns that its new AI Text Classifier tool – like others already available – is not perfect. The method for detecting AI-written writing "will be wrong sometimes," said Jan Leike of OpenAI.

"Because of that," he added, "it shouldn't be solely relied upon when making decisions."

In the U.S.

Teenagers and college students were among the millions of people who began experimenting with ChatGPT after it launched on November 30. The tool is a free service on OpenAI's website.

Many people have found ways to use it creatively and harmlessly. Still, some educators are concerned about the ease with which it could answer take-home test questions or do other assignments.

School districts around the country report they are seeing discussions about ChatGPT change quickly.

By the time schools opened for the new year, New York City, Los Angeles, and other big public school districts in the United States began to block its use in classrooms and on school devices.

The Seattle Public School district blocked ChatGPT on all school devices in December but then opened it to educators. District spokesman Tim Robinson said teachers wanted to use ChatGPT as a teaching tool.

"We can't afford to ignore it," Robinson said.

The district is also discussing expanding the use of ChatGPT to classrooms to let teachers use it to teach students critical thinking. Students could also use the service as a "personal tutor" or to help create ideas when working on an assignment, Robinson said.

OpenAI wrote about the limitations of its detection tool on a blog recently. But the company added that the tool could help to find disinformation campaigns and misuse of AI to mimic humans in addition to catching plagiarism.

The longer a piece of writing, the better the tool is at detecting if an AI system or a human wrote something. AI Text Classifier can examine any piece of writing whether it is a college admissions essay, or a literary study of Ralph Ellison's Invisible Man. The tool will then claim it as either "very unlikely, unlikely, unclear if it is, possibly, or likely" AI-created.

But much like ChatGPT itself, it is not easy to say how AI Text Classifier comes up with a result, Leike said.

There is a lot about the tool that is still not well understood. He said, "There's really not much we could say at this point about how the classifier actually works."

International opinions

Colleges around the world also have begun debating responsible use of AI technology. The Paris Institute of Political Studies, or Sciences Po, one of France's most famous universities, banned its use recently. Sciences Po warned that anyone found using ChatGPT and other AI tools to produce written or spoken work could be banned from the school and other institutions.

To answer criticism, OpenAI said it has been working for several weeks to create new recommendations to help educators.

France's digital economy minister Jean-No235;l Barrot recently met in California with OpenAI leaders, including CEO Sam Altman. Barrot a week later told people gathered at the World Economic Forum in Davos, Switzerland that he was hopeful about the technology. But the government minister said there are also difficult moral questions that will need to be dealt with.

"So if you're in the law faculty, there is room for concern because obviously ChatGPT, among other tools, will be able to deliver exams that are relatively impressive," he said. "If you are in the economics faculty, then you're fine because ChatGPT will have a hard time finding or delivering something that is expected when you are in a graduate-level economics faculty."

He said it will be increasingly important for users to understand how these systems work so they know what biases might exist.

I'm John Russell.

And I'm Ashley Thompson.

Matt O'Brien and Jocelyn Gecker reported on this story for the Associated Press. John Rusell adapted it for VOA Learning English.

_________________________________________________________________

Words in This Story

detect –v. to discover something that was not known or hidden by searching for it

solely – adv. without anything or anyone else involved

district – n. an area established by a government for official government business

tutor – n. a teacher who works with one student

plagiarism – n. the act of using another person's words or ideas without giving credit to that person : the act of plagiarizing something

mimic – v. to copy (someone or someone's behavior, speech, or writing)

graduate – adj. of or relating to a course of studies taken at a college or university after earning a bachelor's degree or other first degree

bias – n. a systematic error introduced by selecting or encouraging one outcome or answer over others


Category