THE NEWS

🤖 The deepfake era is here

In a world full of AI image, video and voice generators you can create almost anything.

With just a handful of photos, the AI can create new pictures of you in places you never went to, doing things you never did. And it’s not just pics. Give an AI a snippet of your voice or a video, and bam! It can fake you saying pretty much anything.

Give these models all the photos and videos of a politician or celebrity and the fakes will be so good, you can't tell them apart from the real deal.

A Wake-Up Call

Last week voters in New Hampshire were getting robocalls with Joe Biden’s voice encouraging them to "save" their vote, noting falsely that, "your vote makes a difference in November, not this Tuesday." The fake voice was created using ElevenLabs, a voice AI tool available to anyone.

These calls were a sneak peek into what we might see more of in this election cycle in the US, if legislation does not come up with new regulations around the technology.

And it's not just about politics. YouTube users in the U.S. found themselves watching scam ads with Jennifer Anniston promoting a too-good-to-be-true Macbook giveaway. Click on the tweet to watch the video. It’s uncanny.

But the real drama kicked off with some AI-generated explicit pics of Taylor Swift that got all over X. One tweet alone got a crazy 47 million views before X finally took it down. Swift's fans were super mad and started posting all their favorite pics of her, trying to get those fakes buried in the search results. At the same time, though, some people were still spreading the fakes around. Things got so out of hand that X had to suspend the search for 'Taylor Swift' altogether.

Deepfake stuff, especially the really dodgy kind with celebrities, used to hide out on the darker corners of the net. However with tools getting easier to use and companies like X and Meta cutting back on human content moderators, it is finding its way to social media.

In response to all this craziness, some lawmakers are getting busy. Congressman Joe Morelle is pushing a law to crack down on deepfake porn. This came up last year after some high school boys in New Jersey made fake explicit images of their classmates using AI. If Morelle’s bill goes through, anyone sharing this kind of content without permission could face major fines and even jail time.

But what about all the scam ads and fake calls? We need some rules for all these AI impersonations, so we don’t start doubting everything we see or hear online.

So, as we dive deeper into this AI world, the big question is: how do we keep the real stuff real? It's a chat we need to have sooner rather than later, with everything getting so mixed up between what's real and what's AI-made.

ONE MORE THING

OpenAI fixes “laziness”

If you've been using ChatGPT for coding, you've probably noticed it's been kinda lazy lately. Instead of giving you the full code, it often just spits out a part and slaps a '#Rest of the code' comment for you to fill in the rest.

Man, there were times I felt like smashing my keyboard! But, OpenAI says they've fixed this with their new GPT-4 Turbo model. It's not available in ChatGPT yet, just through the API, so I haven't had the chance to try it out myself.

One cool update, though – you can now turn on the 'Always expand code output' option in the ChatGPT settings.

⌚ If You Have One More Minute

• 🐇 MORPHEUS-1 The world’s first multi-modal generative ultrasonic transformer designed to induce and stabilize lucid dreams.

•  🎥 Google announced Lumiere, their new AI video model.

• 🇨🇳 China approves over 40 AI models for public use in past six months

• 🚫 What AI can't do at work

AI Art of the Day 🎨

If Red Dead Redemption 2 (game) had a Mexico DLC. Made with Midjourney by u/MrAwesomeJr. Check out all 8 photos here.

🏄🌊🏄🌊🏄🌊🏄🌊🏄🌊🏄🌊🏄🌊🏄🌊🌊🏄🌊🏄🌊🏄

That's all for today, folks!

If you enjoyed this, please consider sharing this hand-crafted newsletter with a friend.