Google CEO Sundar Pichai warned about the dangers of using AI to make fake videos of public figures, saying the task would soon become "easy."
Pichai made the remarks in an interview with CBS' "60 Minutes," telling interviewer Scott Paley that Google was restricting its Bard AI to prevent abuse.
"With artificial intelligence, you know, it's easy to make videos. It could be that Scott said something, or I said something, we never said it," Pichar told Paley. "It may seem accurate. But you know, on a social level, it does a lot of harm. ”
"Is Bud safe for society?" Pelley asked.
"The way we're coming out today, as a limited experiment, I think so. But we must all be responsible for every step forward. ”
"Are you slowly venting so that society can get used to it?" Pelley clarified.
"That's part of it. Part of this is to get feedback from users. Before we build and deploy more robust models, we can develop stronger layers of security. ”
ChatGPT has taken AI by storm in the United States.
Current AI models have been able to create images of public figures that are almost indistinguishable from reality. So far, the production of video and audio is not very proficient.
Google's Bard AI is the company's answer to OpenAI's ChatGPT, which launched late last year and captured viewers around the world.
The audio clips created by AI are capable of mimicking the intended sound, but the current version sounds a bit robotic and unnatural. AI video clips are even less polished.
Pichai went on to say that Google doesn't always fully understand the answers provided by Bard's AI, citing the example of a program that appears to have taught itself Bengali, despite not being trained in it.
"You don't fully understand how it works. And yet, you let society run its course?" Pelley asked.
"Yes. Let me put it this way. I also don't think we fully understand how the human mind works. ”