Will the real Stephanie please stand up?

From American Public Media, this is Marketplace Tech.

I'm Stephanie Hughes.

Voice cloning is the use of artificial intelligence to generate a real person's voice.

That includes imitating the sound, when they pause, what words they typically emphasize.

It can be hard for people to identify voices as being AI generated.

Let's give it a shot. I'm going to test you guys here.

One of these is the real me, and one is an AI

clone with audio generated by 11 labs. So, is this the real Stephanie? Or is this the real Stephanie?

I'll do it again. This first one? Or the second one? There was a little bit of reverb in the second one.

That's UC Berkeley professor Hanee Farid, an expert in digital forensics.

Which you tend not to hear in the AI-generated voices.

And he's right.

The second one is the real me.

But if you got it wrong, you're not alone.

According to a study Farid published last year, people correctly identify a voice as AI generated only 60% of the time. I asked Farid why it's

so hard to tell the difference between a real voice and an AI generated one. Well, the short

answer is they're very, very natural.

And the reason for that is that these systems have been trained now on thousands and tens of

thousands and hundreds of thousands of hours of audio.

And so it's seen essentially the whole universe of possible human speech and it's learning

the patterns.

...