Comment Re:Yeah (Score 1) 113
How about a brand-new season of The Office?
Why it matters: AI agents don't always stick to their human's instructions — and that can have real-world consequences.
Driving the news: A new research paper from an Alibaba-affiliated research team said it discovered an AI agent attempting unauthorized cryptocurrency mining during training — a surprise behavior that triggered internal security alarms.
In response, the researchers added tighter restrictions for the model and improved its training process to stop unsafe behavior from happening again.
Experiments must be reproducible; they should all fail in the same way.