By: Nakya Harris
Artificial Intelligence (AI) voice changer has gained popularity over the past year, creating ways to mimic human speech.
Conversations involving Chat GPT and AI voice have taken over social media. You might have seen students utilizing an online server to write their essays. All it takes is a simple explanation of what is needed and an essay is constructed in seconds. Or, you might have come across the funny viral videos of President Joe Biden saying the funniest phrases. Phrases that he probably should not be saying as president. In reality, that is not “him” saying it. However, it sounds exactly like him. Lastly, scammers have taken advantage of AI voice by tricking individuals into thinking a loved one is asking for money. How? It is the magic of artificial intelligence.
Artificial Intelligence
Hayden Wimmer, an information technology professor at Georgia Southern University, says AI is used to mimic nature or biology on a computer.
When it comes to artificial intelligence, it is a broad term. There are many types of AI.
Wimmer adds that the popular type of AI that we are most familiar with is deep learning. It involves image recognition, image generation, voice changing, and more. This is done in variations of neural networks.
An artificial neural or deep neural network allows computers to shoot for better accuracy. This started with the advancement of hardware and software within the last 10 years. Being able to process images and use math, was a huge step forward for neural networks. Because of the popularity this has gained, more investments have been made in advanced technology.
It is important to note that all types of AI involve trained software to produce specific outcomes.
Artificial Voice Changer
When computers are trained on certain data, it is able to create new data. Since this is brand-new data, it is hard to differentiate whether the output is real or fake. When a computer is trained to recreate an image, it is a totally different image that mimics the original. A couple of years ago, individuals used Photoshop to edit pictures. This does not create a completely new image. Therefore, it is easier to determine the validity. This also applies to AI voice.
“If it’s trying to mimic someone else’s voice, it’s been trained on examples of their voice,” Wimmer says. “So, it’s actually recreating their voice, not changing it.”
There has been worry about this causing future problems, whether on social media, in Hollywood, or everyday life.
Joshua Kennedy, a political science professor, says that even though there are ways to debunk whether audio is AI generated, it will definitely affect politics. Especially when it comes to social media.
"It can have an effect on people's perception of candidates," Kennedy says. "Once that information is out there, it's out there. You can try to prove that it wasn't you, but for some people, it may be too late."
When asked Wimmer what the future holds for AI, he responded, "If I knew where it was going, I'd be rich."