Can you define "AI Safety" more precisely? I could see some grants about the risk of bad outcomes that arent what most people are thinking of when speaking of "AI Safety."

@Jason want to give those examples

@NathanpmYoung Adverse consequences from militarization of AI, where the focus of the grant is not on existential or near-existential risk. That seems somewhere vaguely in between the core "AI ethics" and core "AI safety" camps? Basically, anything where the consequence is non-localized death but is not an existential / near-existential risk could be unclear.