1. 5
  1.  

  2. 2

    I thought this point about communication (on page 13) was especially interesting:

    States frequently are able to communicate their intentions through various forms of signaling, such as explicit threats, arms build-ups, or actual military exercises. For DoD AI applications, many of these assumptions may not hold true. Signaling may be difficult or misperceived at greater rates, knowledge of capabilities or intent is harder to discern, and the source of AI attacks at a distance and in digital form may be harder to attribute.

    Is there any way to make such a communication? Posting job listings, perhaps?