Machine Unlearning, Model Collapse, and the LinkedIn AI Scandal

Congratulations! You've been opted in... to letting LinkedIn clone your posts without crediting you. Recently, LinkedIn opted you in to using your content for training their generative AI systems. Some people are opting out as quickly as their fingers will let them, some are shrugging their indifference, and a few folks are staying opted in on purpose. So let’s unpack what’s going on with LinkedIn OptOut. For bonus points, I'll teach you about about: * data privacy laws * what makes GenAI special when it comes to leaking data * machine unlearning * model collapse (Note: LinkedIn passed over its EU users and initially opted in its UK users until the UK government stepped in. Now EU + UK users are safe and can simply sit back, grab their popcorn, and watch what's happening with the rest of us.) Here's that one-click opt-out link again: https://www.linkedin.com/mypreferences/d/settings/data-for-ai-improvement Machine unlearning review paper: https://arxiv.org/pdf/2405.07406 To join the discussion and to see the poll results showing that 88% of folks who've told me their choice opted OUT: https://bit.ly/linkedinoptinpoll For my more detailed blog posts on the LinkedIn scandal: What happened: https://bit.ly/quaesita_linkedinoptin1 Why it happened: https://bit.ly/quaesita_linkedinoptin2 For technical posts explaining synthetic data and model collapse: About synthetic data: https://bit.ly/quaesita_synthguide2 About AI-generated data: https://bit.ly/quaesita_synthguide4 About model collapse: https://bit.ly/quaesita_synthguide5 ********************************** Invite Cassie to speak at your event: http://makecassietalk.com Decision advising: http://decisionadvising.com Website: http://kozyr.com Intro 1:1s: http://intro.co/cassiekozyrkov