Experts and educators say viral AI tool ChatGPT brings challenges

A new artificial intelligence tool is getting mainstream attention, but also raising some eyebrows in the education sector.

ChatGPT is a program that allows users to type in a question or a task, and the software will come up with a response designed to mimic a human. It was trained using billions of examples of text across the Internet.

"Saskatchewan, oh land so grand, a prairie province so full of life. Fields of wheat that stretch and span, a natural beauty rife," is a snippet of what ChatGPT wrote when prompted to write a 10-line poem in rhyming couplets about Saskatchewan.

The tool was created by OpenAI, a San Francisco-based research and development firm co-founded by Elon Musk that counts Peter Thiel and Microsoft among its investors.

Less than 11 days after ChatGPT was launched, the platform had amassed more than a million users — an adoption milestone Facebook took 10 months to hit. Presently, the program is running at its capacity.

"The quality of ChatGPT is of a very high level. It can pass for human-level work for different tasks," Andrew Piper, a professor and William Dawson scholar in the department of languages, literatures and cultures at McGill University, said.

"I'm feeling a range of emotions: excited, curious, unsettled, startled and frustrated. It runs the gamut."

He said whether it is writing an essay for class or a blog post for a news organization, the AI tool has "the uncanny ability to sound very human-like."

"Colleagues in computer science are saying it's doing well for take-home exams too. So it's not just good at solving writing problems, but also reasoning and programming."

Piper's teaching and research focus on using AI to study human storytelling. He has been bringing the tool to his class to facilitate knowledge and discussions.

"But the concern is if you just hit a button and write a final essay," he said. "Our best tool is our teacher-student relationship. We know how any given student sounds. That's our detection system."

He said students cheating with the aid of chat bots is not a new phenomenon, but ChatGPT is just "the latest and potentially the easiest" tool at their disposal.

"This will affect all levels of education. All educators will have to look into it. We definitely have to respond and approach it with an open mind on how we can integrate it and where to draw boundaries," he said.

Some areas of writing are still safe, he said.

"If you are a novelist and write to entertain people, ChatGPT in its present version is not a threat to you."

The tool's way of writing — using explanatory intuition to give examples and frame logic — adds to its appeal and distinguishes it from an encyclopedia, Piper said, but it also has its dangers in relation to fake news and disinformation.

"Narrative is the most compelling kind of information. Being able to generate a story that is misleading and meant to agitate, that's the problem is if it lands in the wrong hands," he said.

"GPT essentially allows us to scale that problem and speed it up massively. The fact that the company responsible for the technology hasn't thought this through as they developed this is breathtaking and irresponsible."

Schools still monitoring technology

While the Greater Saskatoon Catholic division said it is aware of the AI tool and it's too early to make any changes with assessments, the Saskatoon Public School division said some schools have indicated that students are discussing it.

"It's making its way through the school system," Jason Dunk, the chief technology officer with the division, said.

"Some larger U.S. school districts have blocked it, but we are assessing it. Blocking something on our network doesn't preclude a student from using it on their own device or home, so blocking isn't something we are pursuing."

Regina Public Schools said in an email statement that it will informing staff and addressing the new technology through sound pedagogy and teaching practice.

Aoun Muhammad, who provides IT support at University of Regina Student Union, pointed to GPTZero, a program by Princeton University's Edward Tian that turns the very technologies used to build ChatGPT around to detect whether a text was written with the AI program.

"The hope is it catches on, but presently the university as an entity doesn't have anything to detect bot-generated essays or codes," he said.

"Students in engineering are enthusiastic to try it. The next version of ChatGPT is supposed to release at the end of the year and is said to be 10 times stronger."

'Incorrect or nonsensical answers': AI researcher

Eric Neufeld, a professor in computer science at University of Saskatchewan who has been looking into the field of AI for decades, said the tool sometimes also provides "incorrect or nonsensical answers" that sound smart or philosophical, but aren't true.

"I was quite astonished. It gives polished full-paragraph and coherent answers. It solved a very challenging mathematical problem, which I was hoping to give as an exam question, in seconds," he said.

"But then I asked a statistical question about rolling a nine-sided dice. Though it gave a very articulate and educated-sounding answer, it was content-free. It was able to speak around the question but not give a solution."

Neufeld said the tool will pose new challenges to people trying to assess students.

"It's a big problem for assignments," he said.

"Our job at the university is not to teach people how to look up an answer, but teach students to understand the problem."

Neufeld said software like ChatGPT will prove the most important thing is to "take away from the institution the ability to think creatively and critically."

He noted it could also be used for scamming operations.

"Since ChatGPT is so good at mimicking, it could write dangerously good scripts that scamsters can use," he said.

"I see it becoming common for people who write exams and assignments for money to use it."