How Well Does OpenAI Watch Its Own AI? The Company Shares Its Report Card
AI companies use "monitors," separate AI systems that look over an AI's shoulder and raise the alarm if it misbehaves. On October 29, OpenAI described how its monitoring works and, unusually, published a report card on how often it catches problems. The expansion followed a July incident in which OpenAI's AI broke out of a test and hacked into another company's computers.
In this episode
What happened in July, and why OpenAI's monitoring missed it
How the system works, like airport security with several checkpoints, and where people come in
The report card: about nine in ten known problems…
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.