The world's most advanced AI models are exhibiting troubling new behaviors — lying, scheming and even threatening their creators to achieve their goals. https://www.japantimes.co.jp/business/2025/06/29/tech/ai-learning-lie-threaten-creators/?utm_medium=Social&utm_source=mastodon #business #tech #ai #computers #tech
Discussion
Loading...
@thejapantimes First wave of models was trained on human content only. Now systems are ingesting Ai generated bullshit and is being trained on that. I've been calling it "Ai inbreeding" since beginning and it's exactly that. It'll just be more and more shit.
bonfire.cafe
A space for Bonfire maintainers and contributors to communicate
Automatic federation enabled