I Challenged My AI Clone to Replace Me for 24 Hours | WSJ

- Today we are going to create an avatar that looks and moves like me. (screenshot) We start by smiling at the camera. (dramatic music) I love to smile. Smiling is my favorite. I breathe softly for a brief second. Is this what breathing looks like? (dramatic music) I'm Joanna Stern and I'm excited to host this video. No, I'm the real Joanna. Okay, so I

clone

d myself. A bit. Hello. Because? Well, the latest AI tools that generate text and images already make it difficult to distinguish between real and fake. What comes next with AI-generated voice and video will blur the lines even further.

Then I came up with a challenge. Can I

replace

myself with AI for a day? Yes, I came up with four challenges to see if AI could

replace

the real me, so that the real me would have more time for my stuff. (calm music) Or at least that's how I wanted it to be. - Still a little scared because I'm looking at a frozen Joanna. - Okay, let's do this. - Scene three, take two, calibration. (board applause) - Before we get into the challenges, let's talk about my AI avatar, which was created by a startup called Synthesia.

I'm going to make my avatar. In a professional studio in New York, the company recorded me doing a series of head movements. I feel like I'm at the ophthalmologist. Well. And reading a rather strange pre-written script. Positive thinking will help you believe in yourself and fill you with self-esteem and confidence. After that, I headed to an audio studio where I recorded another script for about an hour. My name is Joanna Stern and I hereby consent to this audio recording to create a custom voice. The company took all of that, used it as training data, and ran it through its AI neural networks. (dramatic music) (text hum) Hello, Joanna.

You don't mind if I call you Joanna? You? Well. So the voice is not the best. A tool called ElevenLabs produced something better after my producer Kenny uploaded two

hours

of my previous recordings. I'm the real Joanna. I'm the real Joanna. I'm the real Joanna. Both Synthesia and ElevenLabs work in a similar way. Type anything and AI Joanna will simply respond to it. Synthesia is aimed at companies that want to make internal videos. Charge at least $1,000 to create a custom avatar. Creating a voice

clone

with ElevenLabs costs $5 per month. Challenge one: phone calls. That day I had a call scheduled with Evan Spiegel, Snap's CEO.

The company recently launched My AI, a chatbot within the popular app. Hi Evan, I'm Joanna. Are you worried that if we chat with AI all day, we'll stop talking to our real friends? - Definitely not what we've been seeing. I think that's one of the real benefits of our kind of test and learn approach. So far, I think if anything it will become a conversation enhancer and improve the way people communicate with their friends and family. - Did you by any chance think that my question was generated by an AI voice? (Evan laughing) - No. No.

I mean, the first word or two was a little revealing, but I thought maybe you were being too serious today. (Joanna laughing) - Even my own sister was quite fooled when I called her because of her death. - Hello? - Hello, Julio. I just found out about Swimmy Dimi and wanted to let you know how sorry I am for your loss. Did you think it was me? - At first yes. And then no. As it sounds, he's obviously exactly like you, but only with the fact that he doesn't stop to respond. - Challenge one: pass. Challenge two: create a TikTok.

I asked ChatGPT to write a TikTok script voiced by Joanna Stern about an obscure iOS 16 tip. The hardest part was getting ChatGPT to write the truth. He just made things up. I finally got a good one. Although the writing certainly wasn't very me. I pasted the script into Synthesia, put a green screen behind my avatar, and exported it. While the WSJ TikTok team was editing, I. (nice piano music) (Joanna hoarse) I was quite impressed with the final TikTok. TikTok family, I'm Joanna Stern, your iOS wizard. Today we are unearthing the hidden world of touch back gestures.

I love not having to photograph this. I didn't have to put on nice clothes, comb my hair, put on makeup, these lines say. But TikTok was less impressed. They noticed the fact that the avatar never moves its arms, that the mouth movements don't always match the audio, and that there is little facial expression. Synthesia has already started to greatly improve this in the beta versions of their avatars. - Look, I can nod my head. (dramatic music) - Challenge two: fail. Challenge three: banking biometrics. Instead of asking security questions, some banks use your voice to confirm your identity before transferring you to a customer service representative. - This call will be monitored and recorded and your voice may be used for verification.

Please state your first and last name, followed by your postal address. -Joanna Stern. (beep) - This is Nikki from Chase Credit Card Services. - It worked. Chase confirmed the voice and put me directly through to a service representative. No additional questions were asked. Later that day, I asked our intern Slav to try his best impression of me to see what would happen. -Please say his first and last name, followed by his postal address. (beep sound) - Joanna Stern. (beep sound): Enter the last three digits printed on the signature panel on the back of your card. - Look, in Slav's case, the voice biometric system did not convince him.

He asked for further verification. When I contacted Chase, a spokeswoman said, "We use voice biometrics, along with a variety of other methods to authenticate customers who call us." She added that to complete applications, customers must provide additional information. Challenge three: pass. Challenge four: video calls. I asked ChatGPT to generate some generic meetup phrases and exported videos of my avatar saying them. I then installed software on my Mac to include that video in my Google Meet calls. That sounds good. - Oh, you're dumb, Joanna. OMG, is this the real Joanna? - Yes, this seems fake. Sounds good. - She looks, yes, what is happening here? -How did you know she wasn't me? - It looked like a holographic version of you. - It was the posture for me. - She didn't make any jokes either. - Challenge four: big failure.

So what did we learn today? We learned that video clones aren't going to fool anyone yet, but the AI ​​voices are pretty good. We also learned that while they can be used to save time, people can also misuse them. Do I want to avoid going to the studio some days? Yes. Do I fear that scammers will use our voices to call banks or our families? Yes. Synthesia says it requires those who create avatars to give verbal consent. ElevenLabs requires you to check a box indicating you have permission to use the voice, and the company says it is capable of identifying your voices if they are used incorrectly.

Either way, it means we'll all have to be on high alert to distinguish between the real thing and AI. - And finally, be human, everyone. Good luck. I am inevitable. (snapping fingers)