Deepfakes, A.I. and you - how the latest tech is helping scammer and what we need to know

FOX 2 (WJBK) - Artificial Intelligence is advancing at the speed of light, making it harder and harder to trust what you see and hear. And the advances in tech make it easy for scammers to take advantage.

A.I. Dave Spencer: "Hi Metro Detroit, you know me, it's Dave Spencer from FOX 2 News. If you are looking to make some quick money, all you have to do is send me a private message indicating your interest.

"What you are seeing and hearing are things I never said or never did, it's all a deepfake."

"That sounded like me," said the real Dave Spencer.

"It did," said David Derigiotis. "It's very easy to pull someone's audio from Facebook, or Instagram, wherever it may be, synthesize it and set up a scam attack."
 
FOX 2 first introduced you to this type of technology back in August.

Meteorologist Derek Kevra used a deepfake version of himself to successfully fool Alan Longstreet and Rich Luterman.

In just six months, Derigiotis says A.I. has come a long way.

"I was able to take a sample of your audio and your video, and run it through a couple of different programs," he said. "Once the voice is able to be cloned and synthesized, I can type whatever I want and make you say whatever I want, and that's what I did with this video."

Merging video with audio adds an extra layer that now needs to be peeled back to make sure you are seeing the truth.

And if this A.I. can be used to separate you from your money, you better believe scammers are going to take advantage.

"The one that you used me for an example, it was kind of a mass scam," said (the real) Dave Spencer. "Just put it out there and see who falls for it."

"They are financially motivated," Derigiotis said. "If you can go after somebody and grab $500 there or $1,000 there - or widespread on someone's social media account that was compromised. Or set up a fake one that looks to mirror yours, you are often able to generate a lot of income off that."

And videos like this can be done relatively quick.

"This video took about a day, I had to train the model with your particular images for the visuals," he said.
 
That's right. Over time, the artificial intelligence will learn to be better.

"We're creating a deepfake video," he said. "So it's the visual of you, this right here, is the original clip, and what you see in the second screen is the artificial intelligence is trying to replicate that."

And practice makes perfect.

"It's gone through 77,000 iterations, so it gets better every single time," he said.

Meaning that we, as consumers, have to be skeptical of everything.

"It's important to verify who you are speaking with, and there are a variety of ways to do just that," said (fake) Dave Spencer.

"I didn't actually say that - but it's the truth. The first thing to do is consider where did the message come from - if it's a video like the virtual Dave Spencer for example. Where did you find it or which account sent it to you?

"You can't just believe a single source, especially if its coming from an account that doesn't have much history behind it," Derigiotis said.

As we inch closer to another election, trusting what we see and hear becomes harder and harder.

Michigan does have laws that require A.I. generated ads that air 90 days before an election to have a disclaimer - but it's still a good idea to use multiple reliable sources before trusting what you see and hear.

The deep fake Dave Spencer, left, and the genuine article on the right.