which has me pondering - how would those elements be best incorporated??

fine tune on psychology exp results that include labels of social/cultural origin, and add real-time levers that modify emotional states, like model temperature but different.

wondering who else is doing this type of training

as a way to reflect back human answers & choice paradigms.

seems to me they would end up with a "WEIRD" model that doesn't take social/cultural factors into account, nor the always ephemeral emotional state & it's influence on cognitive function.

http://llamahub.ai - prebuilt connectors for llamaindex data prep

http://unstructured.io - for embedding documents quickly

http://lmstudio.ai - to connect local models & document repositories