National Security Secret
#177. Realizing an AI we can trust requires raising or training one on proven or trusted knowledge.
Related secrets: #11 #148 #154
Throw the whole internet at a Large Language Model if you feel an insecure, self-important, or biased intelligence has value.
Even worse, amplify that bias through "reinforcement learning" using human feedback.
We'd rather see centuries of common sense put to good use, instead.
Has anyone ever taken time to sort the "wheat from the chaff"?
More information will be provided.
Originally posted on 04-27-2023 @ 01:34 PM
Comments
Post a Comment