Human-like Motivation in LLMs
I think I’m adjusting my take on AI requiring emotional systems similar to those in evolved brains to have intrinsic motivation.
Sentience mimicking language models, trained on a human corpus, could possibly mimic human motivation found there
I think this ‘intrinsic motivation’ can probably be trained out of the models.
My intuition for this current batch of scare-reporting is still that its largely a psyop to protect capital and strategic monopoly
And they still lack another prerequisite for sentience - spontaneous activity