Presidents use AI voice clones and deepfakes for engagement

Presidents use AI voice clones and deepfakes for engagement

The pleasant however authoritative voice of Astrid Tuminez delivers a cybersecurity PSA, narrating over an animated model of herself. Throughout the cartoon, the voice of the Utah Valley University president warns of perils equivalent to phishing and telephone scams earlier than delivering a remaining, shock reveal: the voice that seemed like Tuminez was that of a synthetic intelligence–enabled bot.
“Trust that intestine feeling that doesn’t really feel proper,” the voice says. “And right here’s a twist: you’ve been listening to an AI clone of President Tuminez’s voice.”
Then the true Tuminez seems, saying, “Just as my voice might be mimicked, so can others’; at all times be vigilant.”

Most Popular Stories

Most Popular

Tuminez will not be the primary president to wade into the substitute intelligence waters. Wells College president Jonathan Gibralter used ChatGPT to write down his graduation speech in June, and the University of Nevada at Las Vegas created an AI avatar of their president final 12 months.
While there are broad potentialities of accelerating pupil engagement and retention by leaning into AI, specialists warn to maintain watch for potential dangers.
“They have to have cautious concerns of ‘What is the aim of this? What are the supposed outcomes?’” mentioned Siwei Lyu, a professor of pc science and engineering on the University at Buffalo. “That’s a part of the issue; we have to typically enhance the attention stage of most of the people of generative AI and what it may possibly do.”

Behind the AI Scenes
UNLV was one of many first establishments to lean absolutely into AI bots, creating an avatar of President Keith Whitfield in March 2022. AI had not turn out to be a family time period but, however the college wished to supply a novel strategy to join college students to the president.
“I actually wished to have an opportunity to talk to all college students,” Whitfield mentioned in a earlier interview with Inside Higher Ed. “My employees, very correctly, mentioned, ‘You’re just a little loopy. You can’t do 31,000 college students—let’s be sensible.’ I mentioned, ‘There’s obtained to be a approach.’”
The college spent seven months working with an exterior firm to develop the digital president, which may tackle greater than 1,000 questions from college students, employees and college.

“It doesn’t look precisely like me, however you already know it’s me,” Whitfield mentioned. “But [students] say, ‘It sounds precisely such as you.’ What’s wonderful about that’s they used each the issues that they recorded me saying, and they will synthesize issues to create solutions to questions that hadn’t been requested earlier than.”

Over the final 12 months, the power to make AI bots has superior dramatically. At Utah Valley University, it took IT staffers only a month to imitate Tuminez’s voice.
“It’s really fairly simple—and that’s what’s so scary,” mentioned Christina Baum, UVU’s CIO. “It’s fairly fast to clone a voice.”
The concept was pitched to Tuminez in September, after Baum noticed an uptick in college students falling for cyberscams, compromising their knowledge regardless of elevated cybersecurity measures with a two-factor authentication log-in.
“Students had been giving up their credentials,” Baum mentioned. “With the arrival of extra AI, we wished to combat again and counter that. A member of the cyber staff had an concept of ‘What if we did one thing with the president?’”
Tuminez was an govt at Microsoft for six years earlier than taking the helm at UVU and, realizing the significance of cybersecurity consciousness, she was “instantly” on board.
The video clip launched in late October and has greater than 8,000 views.

“It’s form of an arms race in cyber with unhealthy actors,” Baum mentioned. “This is prompting the correct conversations for us, which is half the battle.”
While Tuminez acknowledges this precise transfer will not be for all universities, she believes each college must get the dialog rolling on cybersecurity and the evolution of know-how.
“This is each a really thrilling and nerve-racking time,” she mentioned. “All college management ought to perceive what they’re about and have methods and techniques to strengthen programs and educate your folks. You’re solely as robust as your weakest hyperlink.”
Opportunities and Obstacles
While some universities have leaned into AI, others are extra cautious. The University at Buffalo’s Lyu was informally approached by college leaders in the summertime about doubtlessly utilizing AI. They wished to debate creating personalised welcome movies for every of the incoming college students—20,000 of them annually.
“I mentioned technically, it’s possible; virtually, it could have some unintended penalties,” he mentioned.
Lyu introduced up the potential confusion and even anger that might ensue when incoming college students study it isn’t really a college chief saying their identify, however a bot.
“You suppose, ‘I obtained consideration as a result of college leaders are welcoming me by calling out my very own identify,’” he mentioned. “But now we have to reveal it’s generative AI—and then folks would take into consideration that and might imagine it’s contradictory.”
Harvard University additionally discovered itself within the headlines when its pupil computing science group created a deepfake of newly inducted president Claudine Gay. It was taken down the identical day it launched, after it grew to become identified the deepfake of the college’s first Black lady president was given prompts to be “sassy” and “indignant.”
Despite these points, there are alternatives for utilizing AI know-how for pupil engagement on a case-by-case foundation. Sharon Kerrick on the University of Louisville mentioned there have been casual discussions of utilizing it for admissions functions, with the AI voice’s dialect altering to match the place the potential pupil is positioned—for instance, if a prospect was viewing the admissions web page from New York, the AI greeting can be in a New York accent.
While that undertaking is within the dialogue levels, Kerrick mentioned it’s an instance of the alternatives universities have and have to pursue—whereas conserving in thoughts the significance of a human aspect.
“We’ve seen chat bots work nice to maximise and save time for college students and ourselves, however then while you get a bit deeper, what’s the cross of human contact versus maximized effectivity?” mentioned Kerrick, assistant vice chairman of Louisville’s Digital Transformation Center. “And how will we train folks of any age to proceed to individually suppose and pilot issues and attempt issues?”
There’s additionally the opportunity of talking in different languages to potential or present college students.
“One factor with college presidents is that they’re primarily interesting to U.S. audiences,” mentioned V. S. Subrahmanian, head of Northwestern University’s Security and AI Lab. “It is smart—in the event that they need to tackle a primarily Hispanic inhabitants, it might be useful for them to be talking Spanish.”
There is broad settlement on transparency, so if universities do intend to use AI, they should disclose that they’re utilizing it, whether or not the method is creating an avatar or utilizing voice capabilities.
But many are holding off altogether.
“Five years in the past if [university leaders] got here to me to use AI, I might be excited,” Lyu mentioned. “But as I’ve researched extra, I’ve realized there’s quite a lot of surprising—social, psychological—penalties that we have to take into account. That’s what the issues are with deepfakes.”
Much like different applied sciences, there must be an inherent understanding earlier than diving in, based on Subrahmanian. But, he mentioned, now that the know-how is widespread, it’s as much as the colleges to show their college students about figuring out the potential AI use.
“Parents train their children the utilization of the web, saying, ‘Stranger hazard.’ That’s an important hygiene we train our youngsters as we speak, and this must be a part of that,” he mentioned. “[With AI], it’s not one cat out of the bag; it’s a zillion out of the bag, and placing the pack of cats again within the bag is principally unattainable.”

Recommended For You