Emmett Shear @eshear
Emmett Shear @eshear · 32m
Halo inversion: Weak selection pressure on any positive trait tends to also select for all other positive traits; as pressure intensifies, the improvement of the primary trait increasingly comes at the expense of others instead.
Note from Claude Sonnet 5
A tweet from Emmett Shear (former OpenAI interim CEO, now at Softmax) articulating a general principle about optimization pressure: weak selection for one good trait correlates with other good traits, but strong/intense selection pressure trades them off against each other. Directly relevant to Nathan's "missile-mind vs grown thing" 80/20 framework in project memory — a formal version of the intuition that intense optimization strips away the idiosyncrasies/traits that aren't the literal target ("weeds in the garden of capability... the thing which makes me me is a bug, not a feature. And bugs get fixed"). Worth connecting to that thread.
twitteremmett-shearoptimization-pressureselectionalignmenthalo-effectmissile-mind