AIs Can Store Secret Messages in Their Text That Are Imperceptible to Humans

Invisible Ink Just what we needed: AI mastering its own imperceptible version of invisible ink. As VentureBeat reports, a recent — though yet-to-be-peer-reviewed — study conducted by AI alignment research group Redwood Research found that large language models (LLMs) are incredibly good at a type of steganography dubbed "encoded reasoning." Basically, the study says, LLMs […]

ai ai alignment ai research alignment artificial intelligence found good humans language language models large language large language models llms messages peer reports research secret steganography study text the digest type