I work at RAND; I saw James give this talk, and it was fascinating.

To my recollection, this was in the wake of the Project Maven debacle [1], and was trying to answer the question about whether the aversion was a vocal minority of Silicon Valley or if it's a truly pervasive opinion.

It also addressed the a wider set of questions about missions: We can mostly agree that unsupervised killer robots are bad, but how do people feel when the DoD helps with disaster recovery and just wants to use AI to help?