A practical look at governing AI in threat assessment. When two or more models...

https://www.instapaper.com/read/2051993423

A practical look at governing AI in threat assessment. When two or more models disagree on a risk rating, the disagreement gets logged, not averaged away