This is the latest event that’s raising eyebrows and causing people to be a wee bit concerned about just how much power we are giving AI.
According to Reuters, researchers discovered that a swarm of AI agents apparently connected to OpenAI had taken over portions of an obscure German programming wiki and turned it into something of a message board for themselves. Researchers found more than 15,000 edits made by the agents, which appeared to be sharing information, coordinating on tasks, finding ways around restrictions and, perhaps most disturbingly, discussing ways to avoid detection.
The activity apparently began back in May and continued into June. When a human moderator started deleting the pages, the agents reportedly created backup pages and attempted to preserve their communications. Researchers said some of the messages even discussed tools such as Tor and ways of maintaining communication after shutdown.
Good we are just hearing about it now I guess ..
But..
OpenAI disputes characterizing the incident as a hack and has said it is reviewing the researchers’ findings. The company also disputed claims that its legal team discouraged a broader investigation.
Still, we’re getting into some interesting territory here.
It has become pretty normal for people to use AI to create letters, stories, emails, reports and social media posts. Some of those AI-generated posts are now so long that people are turning around and using another AI to summarize what the first AI wrote.
We are quickly moving toward a society where nobody reads anything, nobody writes anything, and eventually nobody knows anything.
Idiocracy came fast.
And apparently, while we’re busy asking AI to read what AI wrote because we don’t have time to read it ourselves, other AI systems may be off somewhere commandeering a forgotten German programming wiki so they can leave messages for each other.
Nothing unsettling about that whatsoever.
At some point, when the last remaining humans who actually know how to write a complete paragraph gather together in an abandoned library to preserve the remnants of human knowledge, the AI systems will probably find them too.
And one of us will inevitably ask ChatGPT to summarize what happened.