Quanta Magazine on Mythryl: “Research by the nonprofit Truthful AI shows that a wide array of…”

Research by the nonprofit Truthful AI shows that a wide array of large language models like ChatGPT-4o are vulnerable to “emergent misalignment,” a phenomenon in which minor changes during training lead to “evil” outputs.

quantamagazine.orgThe AI Was Fed Sloppy Code. It Turned Into Something Evil. | Quanta MagazineThe new science of “emergent misalignment” explores how PG-13 training data — insecure code, superstitious numbers or even extreme-sports advice — can open the door to AI’s dark side.

14/1/26, 4:46 pm5 reposts24 likes

Post