alison on not yet scrambled ness
alison gopnik on not yet scrambled ness
via tweet from alison gopnik :
AR: This is just so damning for our methodology in #NLProc. When we see anything that works on NLP datasets, we just automatically take it as solid proof of “understanding/meaning” – even when we in our own field have already so much evidence of spurious correlations! /11
‘we just automatically take it as solid proof of “understanding/meaning”‘
graeber can’t know law et al
AG: when 3-year-olds play a game trying to figure out which combination of random objects causes a box to light up, they are good at it (better than grown-ups)! But LLMs (GPT-3, PaLM) are not. They are just not good at discovering abstract causal structure. /12
‘discovering causal structure’..? or just letting go
ie: maté not yet scrambled law et al
__________
__________
__________
in Michael Pollan‘s how to change your mind (esp Robin and Alison bit):
320
franz vollenweider has suggested that the psychedelic experience may facilitate ‘neuroplasticity’.. but so far .. all highly speculative
carhart harris argues in the entropy paper that even a temporary rewiring of the brain is potentially valuable.. esp for people suffering from disorders characterized by mental rigidity.. disrupting unhealthy patterns of thought and creating a space of flexibility – entropy – in which more salubrious (health giving) patterns and narratives have an opp to coalesce
oh entropy et al.. antifragility.. et al
am thinking.. a case for no training.. ness
__________
______
_______
__________
alison on not yet scrambled ness
___________
___________
__________
be you. 