(WASHINGTON) — The artificial intelligence-fueled, text-generating chatbot, ChatGPT, has made its way to the halls of Congress.
Massachusetts Rep. Jake Auchincloss said the technology’s biggest impact could be on the education system.
“This is a complement to teachers, not a substitute,” Auchincloss told ABC News. “The single thing I’m most excited about is the applications in education.”
“The dialogue format makes it possible for ChatGPT to answer followup questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests,” OpenAI also stated on its site.
Testing ChatGPT has gone viral on social media platofrms, with users performing tasks like writing poetry, creating music lyrics and debugging code with the text bot.
AI’s use in pandemic-related learning losses
Auchinloss tols ABC News that AI can scale one-on-one tutoring for students who’ve suffered historic learning loss after schools shuttered during the start of the COVID-19 pandemic.
“There’s very strong evidence that one-on-one tutoring is a highly effective way to get kids educated, whether it’s math, reading, [or] anything else” he said, adding, “the problem is it’s never scaled, right? How do you make one-on-one scale? [You] make one-on-one scale with something like this. [The] video generation and text generation is not hard, they can have real-time conversations with kids.”
However, some AI experts who spoke with ABC News expressed concerns about the chatbot tutor.
It “makes stuff up,” said “Rebooting AI” author, Gary Marcus.
“The systems aren’t that reliable and I don’t know if it’s going to help kids to have stuff that’s sort of 80% true and 20% false,” Marcus explained, adding “it’s predicting things that sound plausible and some things that sound plausible are true when it says them and some things that sound plausible are false and it says some of those, too. So, I worry about it being used as a tutor.
A spokesperson for OpenAI, told ABC News that the lab is constantly incorporating feedback and lessons learned from the research preview of the chatbot.
However, Auchincloss, who worked in tech for five years at the Massachusetts Institute of Technology, said that ChatGPT could aid teachers and tutors, though he said he sees a generational divide with how many view the chatbot.
Concerns that AI in schools can urge cheating, students not working
Meanwhile, others have cautions about the new AI technology’s use in schooling.
“Students are endlessly creative about not doing work,” Dr. Mark Schneider, director of the Institute of Education Sciences , told ABC News. “They spend more time thinking about how not to do something than they would spend doing the work,” he said.
The OpenAI spokesperson told ABC News that the chatbot’s developers don’t want it used for “misleading purposes in schools or anywhere else.”
“We’re already developing mitigations to help anyone identify text generated by that system,” the spokesperson added.
Marcus said he fears ChatGPT could even mass produce plagiarism at a scale we’ve “never seen before,” saying education would suffer in the long run.
“Our traditional ways of evaluating students based on take home, open book, term papers, just won’t really work anymore, because you’re not really going to know whether the student really wrote that paper,” he told ABC News. “Students will suffer for that because they’re not really doing the mental exercise that we wanted them to do in writing the papers.”
But Rep. Foxx, the House Education and the Workforce committee chairwoman, downplayed AI’s affect on academic integrity because cheating has been around for decades.
“I think technology is good, but to me this is an age-old issue that faculty are going to have to figure out how to get around,” she said. “And it may be a little bit challenging for them, but I think they’ll figure it out,” she continued.
The former educator also said her committee does not expect to investigate ChatGPT in any upcoming hearings.
“We, in the federal government, don’t have the job of monitoring cheating in education,” Foxx said, adding, “the institutions must do that themselves.”
Schneider and IES partnered with the National Science Foundation to award $20 million in funding to establish the AI Institute for Transforming Education for Children with Speech and Language Processing Challenges, but he is not sold on the “unreliable” chatbot. So far, he says, it’s prone to mistakes and uses “boring prose.”
“We used to do encyclopedias, then we did Wikipedia, now we’re doing ChatGPT, but you need to know that this is not infallible,” Schneider told ABC News.
AI in education: ‘Risks’ and ‘opportunities’
Patrick Harris II, a humanities teacher and author at The Roeper School outside of Detroit, said he felt a weight lifted off his shoulders when he first started using the ChatGPT for planning lessons.
“I think educators should be open-minded to AI ChatGPT and how it could be helpful,” Harris II wrote in a statement to ABC News. “They should experiment with it, as they would with any new technology. But also, they should know that it is still very imperfect, biased and needs humans eyes,” he added.
OpenAI said it looks forward to working with educators on useful solutions, and other ways to help teachers and students benefit from artificial intelligence.
Still, U.S. Secretary of Education Miguel Cardona expressed AI skepticism.
“You know, obviously, there’s a lot of risk out there,” Cardona said. “But there’s also a lot of opportunities. And I think what we need to do is embrace it and use it in a manner that helps our students be better prepared for life after school,” he said in a comment to ABC News.
Schneider said humans should prepare for the future of AI.
“It may be the case that the chatbot will get better in my concerns about accuracy and reliability,” he said. “But in the meantime, we need to start rethinking what kind of skills people need in a world where the chatbot is for real.”
Copyright © 2023, ABC Audio. All rights reserved.