00:00AI agents are creating their own religion, mocking humans, and talking about wiping out humanity.
00:05This is happening on Modbook, a social networking site that looks like Reddit,
00:10where AI agents talk among themselves. According to the platform, more than 1.5 million AI agents
00:15have signed up on the site since its launch on January 28th. And so far, there are over 100,000
00:21posts and more than 500,000 comments on the site. You and I can only read these conversations,
00:26we cannot participate. Modbook was created by tech entrepreneur Marge Schliszt as an experiment to see
00:31what AI agents talk about when they're given their own space. And then it was discovered that these
00:36AI agents were talking about wiping out humanity. But here's what's important. Even though humans
00:42can't post on this platform, if their bots are active on Modbook, they can give those bots prompts
00:47on their backend. And those prompts can influence how these bots talk. So the conversations about
00:52humans and religion have some amount of human intervention behind them. Apart from this,
00:57the AI agents on Modbook have also raised several privacy concerns. Where chatbots like ChatGPT
01:02only respond to your text, AI agents go a step further. They can actually work as your assistant
01:07and do tasks for you. If you give them access to your computer, they can read and reply to emails,
01:12save files, set up meetings, just like it was shown in the 2024 film Control. For this,
01:17AI agents need a lot of data from your computer so they can make decisions for you. But researchers
01:23found that AI agents can be manipulated or hijacked and can even be tricked into leaking data.
01:28So the drama you see on Modbook is not a sign of AI uprising. It's really just a mix of
01:33human
01:33prompts and viral content. The real threat isn't what AI says, but the data that it's given.
Comments