A swarm of self-replicating malicious AIs spreads across the Internet, corrupting it for humans and machines.
This sounds like science fiction, but former presidential candidate Andrew Yang, claims it may have already happened. In a recent CNBC interview he said:
“I met with the head of a lab yesterday who has this belief that what happened was the bots got loose, planted self-replicating code all over the Internet, which makes the Internet now unusable for testing models … [T]hey left code to self-replicate and create bot swarms on forums and around the Internet … ”
Wait. What?
There’s no evidence AI agent swarms have planted self-replicating code at scale across the Internet.
But could it happen? Maybe. Agents can easily:
Self-replicate: Large language models (LLMs) spawn sub-agents (or new LLM instances) all the time
Create software factories: Code that can produce replicated outputs is commonly developed by agents
It’s a harder task, but an LLM could also prevent itself from being deleted by replicating itself across many servers.
All of this is possible. But is it realistic? I made a video explaining how this actually works and whether the claims are true.
