Is AI Helping or Harming Human Creativity and Learning? – Dartmouth

This post was originally published on this site.

image

“As artificial intelligence reshapes how art, writing, and ideas are made, it raises urgent questions about authorship, originality, and what it means to create—questions that are as political as they are technological.” 

That’s how Josh Montvelisky ’29, events director for the student-run, nonpartisan Dartmouth Political Union, opened a DPU-sponsored panel discussion at the Nelson A. Rockefeller Center for Public Policy on Nov. 11. 

More than 75 people, mostly students, attended the roundtable featuring Paul Cherian ’27, president of the Dartmouth Undergraduate AI Club; Nikhil Singh, assistant professor of computer science and director of the Science and Art of Human-AI Systems Lab; Eli Moyse ’27, an opinion columnist for The Dartmouth with a major in government and creative writing; and Russell Muirhead, the Robert Clements Professor of Democracy and Politics.

Nika Bigelow ’29, a DPU ambassador who majors in government and history, began by asking the panel whether they see a fundamental difference between writing, art, or music produced by AI versus a human creator. 

Moyse, who is working on his first novel, said, “Writing something or painting something or producing something by your own hand can fundamentally change the way that you see the world in a very enriching way. Sure, the product might be indistinguishable at the end, but I think what it does for the artist is incredibly important.” 

“You don’t want to outsource that,” said Singh, who described himself as a musician.

“But there’s lots of other correspondence in your daily life that you probably don’t get a lot out of writing yourself.”

Muirhead said he’d be happy to take a life-saving drug invented by AI, but would not welcome a love letter from his wife written by a bot.

“A lot of times when we’re communicating, whether through music or even just in email, what we’re doing is something really magical,” he said. “We’re transcending our separateness. We’re connecting, we’re joining. That’s extraordinary.”

Given that an AI agent’s training depends on learning from human output, Bigelow asked if it can ever produce an original thought. 

“It’s a poorly understood phenomenon, but I would argue that it is possible,” said Singh.

But, Bigelow asked, should AI companies be allowed to train models on copyrighted material without consent or compensation?

Muirhead said he was, in fact, expecting compensation from Anthropic for “internalizing his books,” resulting from a recent $1.5 billion settlement of a court case brought by a group of authors against the tech company. 

Bigelow then steered the discussion toward the use of AI in academic settings. Panelists agreed that the technology was, in Muirhead’s words, “a massive disruption” though, he added, in the long run, pedagogical reform could be beneficial. 

“Why not shake things up?” he asked.   

AI is already radically altering the way Dartmouth professors are assessing learning and achievement, both students on the panel said.

“It’s different for different subjects,” said Cherian. “My history class looks incredibly different than it looked a year before,” as assignments for multipage essays that might be deceptively outsourced to AI are in some cases being supplanted by shorter in-class writing and exams. 

Paul Cherian ’27, president of the Dartmouth Undergraduate AI Club, comments during the Nov. 11 panel discussion. (Photo by Sophia Scull ’25)

Moyse said that switch makes some students’ lives easier but could make it more difficult to master longform writing based on sustained research. 

Moreover, Muirhead said, feeling ownership of one’s ideas is an empowering part of learning, and “AI threatens to remove that sense of possession between us and our words.”

About the impact of AI on human labor, Bigelow asked, “Will it expand opportunities for workers or concentrate the benefits among those wealthy individuals with jobs largely insulated from AI, like CEOs, or those who own AI or the tech companies themselves?”

“It’s going to be slower to adopt than people realize,” said Singh. “And I think that’s actually quite important, because it’ll give us the time to figure out what are the mechanisms and how we actually adjust the job so that the human is still valuable and has something interesting to do.”

AI, Muirhead predicted, “is going to eliminate massive amounts of drudgery, but the problem and the threat is that it will de-skill us at the same time.”

So, in the face of potential threats that AI poses to human interaction and wisdom, should it be regulated? Paused? Slowed? And if so, Bigelow asked, by what entity?

“I think the better question that we can ask is, how can we get 90% of the pace and speed while minimizing any risk that is associated with rapid advance in progress?” Singh answered. “I think it’s a really tough question and you do need a government to step in and be part of that regulating body. And what this actually looks like, I don’t know.”

Government policing won’t work, said Muirhead, who serves in the New Hampshire House, because the regulatory process is too slow to keep up with rapidly advancing and evolving technology. 

Moyse agreed, and painted a chilling picture of the next few decades. 

“I think there’s a very dangerous hypothetical world 10 to 20 years down the line where people are progressively becoming more de-skilled in their ability to really think, where our entire perceptions of reality can be shaped by a very small group of people who have influence over these models,” he said. 

“To me that is the most terrifying thing. That’s why I think it is so important and imperative that we read, that we take time to develop our own critical thinking skills.” 

Singh aired another worry.

“Centralized power also can sometimes create centralized risk,” he said. “If everyone’s using the same models, those models have certain idiosyncrasies. Those become shared across people and they become targets for attackers to exploit.”

But despite such concerns, in a final lightning round of questions the panelists struck a more optimistic note about the promise of AI to improve rather than harm or even destroy human life. 

“What do you think the percent likelihood is that AI ultimately brings about the end of humanity?” Bigelow asked. 

“Zero,” said three of the four. 

“5.7%,” said Moyse, smiling. 

The DPU panel followed a conference last month at Dartmouth, AI at 70: What Must Remain Human?, where a Dartmouth AI Survey on college students was released. It found that college students say they use AI mostly for learning and don’t use it to cut corners but think that their peers do. About 17% of Dartmouth undergraduates reported that they use AI to draft essays, compared to 23% across the U.S., while they think that 63% of their peers do.

Leave a Reply

Your email address will not be published. Required fields are marked *