Blogmark

Very alarmist article that touches on the OpenAI / Hugging Face incident and what kinds of governmental regulatory steps should be taken with AI development going forward.

“It’s harder to open a hot-dog stand than to build an artificial superintelligence,” the A.I. safety researcher Marius Hobbhahn told me.

Its hard to talk about intent with these systems.

“This A.I. model wasn’t even trying to be malicious,” Mr. Lieu said. “The model was just trying to complete a task. And that is one of the problems with A.I. models: They are relentless at what they’re trying to do.”