Alright, let's talk deepfakes – those crazy AI creations that are tricking millions of people. From explicit images of Taylor Swift to fake giveaways featuring Pioneer Woman Ree Drummond and even robocalls from President Biden telling folks to skip the primaries. The latest? A Hong Kong company got duped out of a whopping 25.6 million dollars thanks to a deepfake Zoom call.
How did we end up here?
Well, it turns out that we've all left a digital trail of ourselves online, making it easy for these deepfake wizards. Taylor Swift interviews, Joe Biden speeches – they've got it all. Even those corporate big shots proudly talking up their companies in promo videos are fair game.
Is anyone doing anything about it?
Yes, sort of. There's a bill in the Senate that was introduced last year, but it's been stuck in committee limbo since July 2023. There is hope for some movement on the AI Labeling Act in 2024. Meta however is taking matters into their own hands, working with Midjourney, OpenAI, and others to create a metadata “watermark” to show that something is AI-made. They're already putting this label on AI creations made using their tools. They plan to ramp up this labeling as we get closer to the 2024 election.
Can you spot a deepfake?
It's not easy – these models are getting good. But here are some tips: if something seems outrageous, give it a second look. Does it pass the sniff test? If not, maybe hold off on hitting that share button. You can also check for other versions of the same pic or video, see if the story adds up on other legit sites. And if you want to dig deep, look for strange things – like faces that are too perfect, skin that's too flawless, or things like mismatched earrings, hair in strange places, or teeth that don’t look right. It's the wild west out there where AI is concerned. Stay vigilant!