Why generative AI 'hallucinates' and makes up stuff

University of Rochester’s Christopher Kanan says current iterations of AI lack human-like self-awareness and reasoning abilities.

Apr 10, 2025

2 min

Christopher Kanan

Generative artificial intelligence tools, like OpenAI’s GPT-4, are sometimes full of bunk.


Yes, they excel at tasks involving human language, like translating, writing essays, and acting as a personalized writing tutor. They even ace standardized tests. And they’re rapidly improving.


But they also “hallucinate,” which is the term scientists use to describe when AI tools produce information that sounds plausible but is incorrect. Worse, they do so with such confidence that their errors are sometimes difficult to spot.


Christopher Kanan, an associate professor of computer science with an appointment at the Goergen Institute for Data Science and Artificial Intelligence at the University of Rochester, explains that the reasoning and planning capabilities of AI tools are still limited compared with those of humans, who excel at continual learning.


“They don’t continually learn from experience,” Kanan says of AI tools. “Their knowledge is effectively frozen after training, meaning they lack awareness of recent developments or ongoing changes in the world.”


Current generative AI systems also lack what’s known as metacognition.


“That means they typically don’t know what they don’t know, and they rarely ask clarifying questions when faced with uncertainty or ambiguous prompts,” Kanan says. “This absence of self-awareness limits their effectiveness in real-world interactions.”


Kanan is an expert in artificial intelligence, continual learning, and brain-inspired algorithms who welcomes inquiries from journalists and knowledge seekers. He recently shared his thoughts on AI with WAMC Northeast Public Radio and with the University of Rochester News Center.


Reach out to Kanan by clicking on his profile.


Connect with:
Christopher Kanan

Christopher Kanan

Associate Professor of Computer Science

Christopher Kanan's research focuses on deep learning and Artificial Intelligence (AI)

AI and Machine LearningApplied Machine Learning (e.g. Medical Computer Vision)Language-guided Scene UnderstandingArtificial IntelligenceDeep Learning
Powered by

You might also like...

Check out some other posts from University of Rochester

Target Can’t Seem to Escape the Crosshairs featured image

1 min

Target Can’t Seem to Escape the Crosshairs

The on-again-off-again nationwide boycott of Target has the retailer’s new chief executive, Michael Fiddelke, officer facing relentless pressure from activists on both sides of the issue. David Primo, a professor of political science and business administration at the University of Rochester, says Fiddelke can’t seem to move Target from the crosshairs despite slashing prices on thousands of products and investing in stores, workers, and technology. “Target remains a battleground for activists on the left and the right, and its new CEO hasn’t yet figured out how to extricate the company from this role,” Primo recently told USA Today. “Fiddelke already faces a huge challenge in turning around a company with significant operational issues. This certainly doesn’t help matters.” Target has reported 13 straight quarters of sluggish sales. Company officials have admitted that shopper anger has contributed. Activists in Minneapolis, where Target is based, organized a nationwide boycott last year over the company’s rollback of diversity, equity, and inclusion policies. From church pulpits to community gatherings, the policy about-face was widely viewed as a betrayal of Black Americans who had propped up the retail giant’s bottom line. Primo studies corporate political strategies, among other areas, and regularly shares his insights with business journalists and political reporters. His essays have appeared in The New York Times and The Wall Street Journal, and he’s been interviewed by many radio and television outlets, including Bloomberg and National Public Radio. Contact him by clicking on his profile.

Fewer Parents are Reading to Their Kids—and Why It Matters featured image

2 min

Fewer Parents are Reading to Their Kids—and Why It Matters

A dramatic decline in reading for pleasure in the United States has fewer American parents reading aloud to their children — and experts warn the consequences can be dire. “It builds connections,” Carol Anne St. George, an expert in early literacy at the University of Rochester’s Warner School of Education and Human Development, recently told The74 for an article citing a 41-percent decline in parents reading to children daily. “People talk about text to text, text to world,” St. George said, “and those are the kinds of things that help children cognitively think and classify their world around them.” Many young parents grew up in an education system focused on reading as a means to testing and building skills rather than enjoyment. As a result, St. George worries, they often view reading to their young as an obligation rather than a joy and a time to bond. Experts say an increased reliance on screens and digital content and time pressures and competing demands on families have also fueled the decline. St. George notes that children benefit greatly from being read to regularly. The advantages of early literacy include: • Having a more robust vocabulary and stronger communications skills. • Being better prepared to learn in school. • Having a closer relationship with their parents. • Higher academic achievement and better health outcomes later in life. What Parents Can Do St. George advises parents to: • Let children choose books they enjoy. • Make reading part of a daily routine and that bedtime is ideal. • Focus on fun and connection. • Model good reading behavior because children mimic what they see. St. George is available for media interviews and can be reached by contacting Theresa Danylak, the director of communications at the Warner School, at tdanylak@warner.rochester.edu.

The truth behind federal disclosure of alien life featured image

1 min

The truth behind federal disclosure of alien life

With the recent presidential comments on potential alien life, UFO enthusiasts have new hope that finally we’re going to get federal “disclosure” of UFOs, aliens and the great government conspiracy surrounding both. But, as a scientist who studies the search for life in the Universe, the question I have is much simpler: What would disclosure really need to disclose? What is required for actual, factual proof that aliens exist and they’ve been visiting Earth? We’ve already had three years of Congressional hearings on UFOs that have produced zero proof of anything. What we need now is simple: hard physical evidence. That is what disclosure needs to deliver. Not stories about alien spaceships being held by the government, but the actual spaceships themselves. Not stories about alien bodies but the actual icky, gooey bodies with their icky gooey tentacles. If disclosure provides physical evidence that independent laboratories and independent scientists all over the world can verify, then it will live up to its hype. That would make “Disclosure Day” truly history-making.

View all posts