Applications
Intelligence services
organization's digital personae (identity)
UpdateEnsemblePersonality
the character of Zcash communities
hidden state
geo, audiences
market intelligence
"secret contracts," it also facilitates "coin mixing"
know the communities
risk prediction
NN, ARNN augmented by context
ensambles
model trustability (complexity, not perfect predictability)
reason about interventions: what if I change X
do not create models vs create models dynamically, spatio-temporal encoding of information
fake news
alternative facts. provenance of the data, context, structure
deepfakes ( no single arbiter of truth , public)
genome (trees)
breakdown of symmetry
humans spread pleasant-sounding lies
rule of thumb, credulity
cognitive dissonance : double down on beliefs
taxes?
Conclusions
Cultural algos
next
to reason about introspection: fantasy
intuition about cause and effect
a model of the environment
Irreversibility . the blockchain is irreversible social computing \cite{rao}.
solve vulnerability (Moshe Y. Vardi), do the job, do not lose money
trustable symbolic regression
trustworthiness, transparency
AI belief (update "mental" model)
blind trust vs earn trust (Kerstin Vignard), verifiable answers (Manuela Veloso) consistency of answer
the question is not if we should trust AIs, but rather if AIs should trust us
deal with edge cases