I'd say (1) is authentically grappling with a concept based on one's own experience, and (2) is mindless mimesis
You're right humans do both (1) and (2) and we are to an extent "trained" by outside info. But our model weights (or brain equivalent thereof) are also constantly tuned by our own experience which leads us to grapple with things. AI agents might *sound* like they are grappling when they're actually just mindlessly outputting next tokens as predicted by training data.
But it's hard to tell the difference
This has finally become an audiobook. It's been on my to-read list since 2018 when I bought a paper copy -- the only format available until now -- but have procrastinated reading because the topic feels heavy.
Looking forward to binging the audio version
"Every Elephant Has Its Own Name, Study Suggests"
An analysis of elephant calls using an artificial intelligence tool suggests that the animals *may use and respond to individualized rumbles.*
https://t.co/lp1TDbjoeT