You have /5 articles left.
Sign up for a free account or log in.

A male student on the left is speaking. On the right a phone displays the new ChatGPT AI. Speaking on both sides is indicated by speech bubbles containing icons representing academic subjects.

ChatGPT’s newest version, GPT-4o, aims to converse aloud in a more human way, leading educators to consider new applications for teaching and learning.

Photo illustration by Justin Morrison/Inside Higher Ed | 4x6/E+/Getty Images | Vladislav Popov/iStock/Getty Images | Chotika/rawpixel | OpenAI

Haya Ajjan eagerly sat in front of her computer on Monday, joking she was on the edge of her seat, watching OpenAI announce its newest iteration of ChatGPT.

“I thought, right away, this is going to change personalized learning,” said Ajjan, associate dean at Elon University’s Love School of Business in North Carolina.

ChatGPT’s newest version, GPT-4o ( the “o” standing for “omni,” meaning “all”), has a more realistic voice and quicker verbal response time, both aiming to sound more human. The version, which should be available to free ChatGPT users in coming weeks—a change also hailed by educators—allows people to interrupt it while it speaks, simulates more emotions with its voice and translates languages in real time. It also can understand instructions in text and images and has improved video capabilities.

The new version adds to the tsunami of interest in generative artificial intelligence since ChatGPT’s launch in Nov. 2022. Over the last two years, some in higher education have shunned AI, while others embraced it, and the majority have cautiously begun tinkering.

Ajjan said she immediately thought the new vocal and video capabilities could allow GPT to serve as a personalized tutor. Personalized learning has been a focus for educators grappling with the looming enrollment cliff and for those pushing for student success.

“That application in education is really profound,” Ajjan said. She gave the example of solving a math problem, where GPT could walk a student through solving the equation.

With its announcement, OpenAI released a video showing that very example.

“The technology would allow us to lift up the learning curve,” Ajjan said, describing the Bloom’s 2 Sigma Problem, which theorizes students that get one-on-one tutoring perform better than those in a standard classroom environment. “That problem would be resolved; this is a gamechanger.”

Balancing Concerns and Benefits

As with all things at the intersection of AI and education, the latest advance was also greeted with caution.

“Despite the advancements of GPT-4o, it’s important to remember that GenAI tools should be used to augment, not replace, the traditional learning experience,” said Darren Person, chief digital officer at Cengage Group, an education and technology firm in Boston. “As the education community experiments with these tools, it is imperative that the safety and privacy of students and faculty are not put at risk.”

While praising the recent advances for the potential personalized learning and tutor support, Person said tech companies release these models to capture data to improve the technology, so “it’s important to be cautious, as adopting these models too early can pose a significant risk.”

Risks aside, Gašper Beguš, director of the speech and computation lab at University of California, Berkeley, said the new technology could help supplement learning by delving deeper into specific concepts.

“When you have a book and don’t understand a passage, you have to ask someone,” Beguš said. “With this, if you don’t understand a subsection, [GPT] can answer you. It’s super powerful; the future is in the very personalized curricula they will be able to offer.”

There’s also the potential for role playing, according to Ajjan. She pointed to mock interviews students could do to prepare for job interviews, or, for example, using GPT to play the role of a buyer to help prepare students in an economics course.

AIs That Listen and Speak

GPT-4o joins a growing number of offerings from tech companies focused on verbal communication. Those include Google’s Gemini Live, announced Tuesday, which lets users have real-time conversations with the chatbot. OpenAI also has had earlier forays into voice responses: premium and enterprise models of ChatGPT already offered it, with five types of voices.

Terumi Miyazoe, a senior associate professor at Tokyo University of Science, used a previous form of GPT’s voice-interaction features last fall to create a lesson design. She said it was “comparable to a colleague.”

In a case study she conducted using the voice interaction features with five students, she found students asked the bot more challenging questions, something they may not have done with their professor.

Ajjan agreed, saying the tool could also supplement questions otherwise asked during office hours.

“What will happen is students will feel more comfortable to ask questions they didn’t want to in class, and dive deeper into the larger subject I’m teaching,” she said.

Beguš said that since the voice of the new ChatGPT sounds more human, the tool could see a bigger uptick in use compared to a system like Apple’s Siri, which he said sounds robotic. That said, he does not believe the talking AIs will serve as replacements for teachers.

“You can ask ChatGPT about anything you want to learn about, but you need the inspiration to learn in the first place,” he said. “I think that’s what we go to humans for. You want people who are inspired by an object of study and have a human conversation.”

He and Ajjan both said it is important to continue evaluating the technology and its potential pitfalls — but to continue experimenting.

“This technology is going to get better with every version, but it’s important to understand the limitations while still experimenting,” Ajjan said. “We can’t shy away from it. Even the faculty who say, ‘I don’t know where to start,’ or ‘This doesn’t impact my field,’—I don’t think that’s a thing anymore. OpenAI opened the entry field by removing the cost associated with it.”

GPT-4o allows all users to access GPT-4. The older GPT-3.5 served as the previous free model. The new version has fewer false facts—otherwise known as hallucinations—and it is quicker and more nuanced. That could push more professors to utilize the technology, according to Ethan Mollick, an associate professor at the Wharton School of the University of Pennsylvania and an AI analyst.

“[The] biggest actual implication of today’s OpenAI announcement is very practical: The top barrier I see when I give talks on using AI is that people don’t pay for AI to start, and they use GPT-3.5 (the free model) and are disappointed, not knowing that the GPT-4 model is 10 times as good,” he said in a LinkedIn post. “Now everyone around the world gets GPT-4 free. This also has huge implications for education and other uses where equity of access to top models has been an ongoing concern.”

Next Story

Written By

More from Artificial Intelligence