Latest / Privacy Please / S6, E247 - The EU's Bold Move on AI Training

Transcript

S6, E247 - The EU's Bold Move on AI Training · Privacy Please

• 0:00
  Speaker 1: Is your favorite AI creating art from stolen goods?
• 0:03
  Speaker 1: It's a huge question nobody was ready for, but Europe is tackling it head on.
• 0:09
  Speaker 1: Welcome to Privacy, please, where we decode the digital world for you.
• 0:13
  Speaker 1: I'm your host, cameron Ivey, and today we're diving into a bombshell 175-page study from the European Parliament that could change the future of AI forever.
• 0:25
  Speaker 1: So you know those amazing AI tools that can write a poem, design a logo or even generate a photorealistic image of your dog on a moon.
• 0:42
  Speaker 1: They seem like magic, but here's the not-so-magical secret.
• 0:47
  Speaker 1: To learn how to do that, they have to study A lot.
• 0:51
  Speaker 1: And what are they studying?
• 0:54
  Speaker 1: The entire internet that includes your favorite artist's profile portfolio, the photographer you follow on Instagram, best-selling novels and investigative journalism.
• 1:05
  Speaker 1: They are, in essence, reading everything.
• 1:10
  Speaker 1: The big problem they're not asking for permission.
• 1:15
  Speaker 1: This new EU study basically says hold on a second.
• 1:18
  Speaker 1: They've concluded that the way these AI models are trained on copyrighted material is a form of mass reproduction, and the current laws just aren't built for this.
• 1:30
  Speaker 1: Just imagine a robot grabbing books and art and photos from a digital library.
• 1:36
  Speaker 1: So what did the EU's deep dive uncover?
• 1:41
  Speaker 1: 175 pages is a lot.
• 1:43
  Speaker 1: If you don't have the time, I'm going to break it down for you into three main bombshells.
• 1:47
  Speaker 1: Number one the all-you-can-eat buffet is over.
• 1:51
  Speaker 1: Right now, ai companies are essentially treating the internet like a free all-you-can-eat buffet.
• 1:57
  Speaker 1: The EU has a rule called the text and data mining, or TDM, which was meant for research, but this study shows training a massive commercial AI model is not the same as academic research.
• 2:13
  Speaker 1: The big recommendation flipping the script from an opt out system to an opt in.
• 2:18
  Speaker 1: Think of it like this Opt out is like a restaurant automatically adding a 20% tip to your bill and you have to notice it and ask to have it removed.
• 2:27
  Speaker 1: Opt-in is when they ask you first if you'd like to leave a tip.
• 2:31
  Speaker 1: The EU is suggesting that AI companies should have to ask for permission to use creative work for training, not just take it until someone tells them to stop.
• 2:42
  Speaker 1: This is a game changer for creators.
• 2:46
  Speaker 1: Number two we need to see the receipts Right now.
• 2:50
  Speaker 1: What these AI models are trained on is a total black box.
• 2:54
  Speaker 1: It's a secret recipe, but what if that recipe includes your private photos, biased information or pirated content?
• 3:03
  Speaker 1: The study demands radical transparency.
• 3:05
  Speaker 1: It proposes that AI companies must provide a detailed summary of what their models have been trained on.
• 3:11
  Speaker 1: This isn't just about copyright.
• 3:13
  Speaker 1: It's a massive win for privacy and fighting bias.
• 3:17
  Speaker 1: It means we could finally see if our personal data is being used to train the next big AI.
• 3:25
  Speaker 1: And finally, number three, fair pay for fair play.
• 3:28
  Speaker 1: If an AI learns its style from thousands of artists, shouldn't those artists get a piece of the pie?
• 3:35
  Speaker 1: The study says a resounding yes.
• 3:38
  Speaker 1: It calls for a new and fair licensing models.
• 3:41
  Speaker 1: Imagine a system, maybe like Spotify for AI training, where creators get paid every time their work is used to teach a machine.
• 3:49
  Speaker 1: This could ensure that humans who create the foundational culture and knowledge of AI are actually compensated.
• 3:57
  Speaker 1: Weird.
• 3:59
  Speaker 1: So what does this all mean for you on your phone, right now, on your computer?
• 4:03
  Speaker 1: This is more than just a legal document.
• 4:05
  Speaker 1: It's a battle for the future of creativity and, yes, your privacy.
• 4:10
  Speaker 1: The EU's AI Act is already putting some of these ideas into motion, with new rules set to take effect in August.
• 4:17
  Speaker 1: This study will fuel the fire for even stronger protections.
• 4:21
  Speaker 1: It pushes for a world where AI innovation doesn't come at a cost for human creators.
• 4:26
  Speaker 1: It's a call for an AI that is not only smart but also fair and transparent, and in a world where our digital lives are constantly being mined for data.
• 4:36
  Speaker 1: That's a privacy conversation we all need to be a part of, so next time you ask an AI to create something for you, think about what it learned from Thanks to the EU.
• 4:47
  Speaker 1: The answer to that question might soon be public knowledge.
• 4:51
  Speaker 1: That's all the time we have for this episode of Privacy.
• 4:54
  Speaker 1: Please.
• 4:54
  Speaker 1: I'm Cameron Ivey.
• 4:56
  Speaker 1: Stay safe, stay informed and always question the code.