A New AI Chatbot Might Do Your Homework For You. But It's Still Not An A+ Student

Why do your homework when a chatbot can do it for you? A new artificial intelligence tool called ChatGPT has excited the internet with its superhuman abilities to solve math problems, write college essays and write research papers.

After developer OpenAI released the text-based system to the public last month, some educators raised alarms about the potential for such AI systems to transform academia, for better and worse.

“AI has basically ruined homework,” said Ethan Mollick, a professor at the University of Pennsylvania’s Wharton Business School, on Twitter.

The instrument was an instant hit with many of his students, he told NPR in an interview Morning editionwith its most immediate use being a way to cheat by plagiarizing the AI-written work, he said.

Aside from academic cheating, Mollick also sees its benefits as a learning partner.

He used it as his own teaching assistant for help in creating a syllabus, lecture, an assignment and a grading rubric for MBA students.

“You can paste in whole academic papers and ask it to summarize it. You can ask it to find an error in your code and fix it and tell you why you got it wrong,” he said . “It’s this multiplier of capability, which I think we don’t quite get our heads around, it’s absolutely amazing,” he said.

A convincing – yet unreliable – bot

But the superhuman virtual assistant — like any emerging AI technology — has its limitations. After all, ChatGPT was created by people. OpenAI trained the tool using a large dataset of real human conversations.

“The best way to think about this is that you’re talking to an all-knowing, eager-to-please internal who sometimes lies to you,” Mollick said.

It also lies in self-confidence. Despite its authoritative tone, there have been cases where ChatGPT won’t tell you when it doesn’t have the answer.

That’s what Teresa Kubacka, a data scientist based in Zurich, Switzerland, found when she experimented with the language model. Kubacka, who studied physics for her Ph.D. studied, tested the tool by asking it about a made-up physical phenomenon.

“I deliberately asked it about something that I thought I knew didn’t exist so that they could judge whether it actually also has the idea of what exists and what doesn’t exist,” she said.

ChatGPT produced an answer that sounded so specific and plausible, backed up with citations, she said, that she had to investigate whether the bogus phenomenon, “a cycloidal inverted electromagnon,” was actually real.

When she looked closer, the alleged source material was also fake, she said. There were names of well-known physics experts listed – however, the titles of the publications they supposedly wrote were non-existent, she said.

“That’s where it kind of gets dangerous,” Kubacka said. “The moment you can’t trust the references, that also erodes the confidence to cite science at all,” she said.

Scientists call these false generations “hallucinations.”

“There are still many cases where you ask it a question and it will give you a very impressive-sounding answer that is just dead wrong,” says Oren Etzioni, the founding CEO of the Allen Institute for AI, which runs the research nonprofit until recently. “And it’s obviously a problem if you don’t carefully verify or corroborate the facts.”

Users experimenting with the chatbot are warned before testing the tool ChatGPT "may sometimes generate incorrect or misleading information." — Users experimenting with the chatbot are warned before testing the tool that ChatGPT “may sometimes generate incorrect or misleading information.”

An opportunity to explore AI language tools

Users experimenting with the chatbot’s free preview are warned before testing the tool that ChatGPT “may sometimes generate incorrect or misleading information,” harmful instructions or biased content.

Sam Altman, OpenAI’s CEO, said earlier this month that it would be a mistake to rely on the tool for anything “important” in its current iteration. “This is a preview of progress,” he tweeted.

The flaws of another AI language model revealed by Meta last month led to its shutdown. The company pulled its demo for Galactica, a tool designed to help scientists, just three days after encouraging the public to test it out, following criticism that it spewed out biased and nonsensical text.

Etzioni also says that ChatGPT does not produce good science. For all its flaws, however, he sees ChatGPT’s public debut as a positive. He sees it as a moment for peer review.

“ChatGPT is only a few days old, I’d like to say,” said Etzioni, who remains at the AI Institute as a board member and advisor. This gives us a chance to understand what he can and cannot do and to seriously start the conversation of ‘What are we going to do about it?’ ”

The alternative, which he describes as “security through obscurity,” won’t help improve fallible AI, he said. “What if we hide the problems? Would that be a recipe for solving them? Typically—not in the world of software—that hasn’t worked out.”