the framework of trying to find a formula might be a trap. Cognitive work might be fundamentally discrete, path-dependent and irreducibly multi-dimensional in a way that resists equations. Or is it?
Discussion
as a wordmetician, i concur :)
one should not & can not
mathmetize e v e r y t h i n g❕
math doesn't need to be the goal in order to be useful. by trying to write a formula, we have to think about what the terms are; what units we use; the scaling factors etc.
is the fundamental unit of progress? words / lines of code / constraints / axioms / frameworks /
and efficient? alignment / synchronization
and what units do we use for ai? cpu hours / parameter counts / tokens
given all of this, can we say that constraints are discovered at a certain pace relative to tokens of synthetic intelligence?
that frameworks of a certain complexity require a minimum parameter count?
to circle back, it would seem that synthetic intelligence is a tool that converts energy into conceptual progress
with a potentially measurable efficiency
v challenging stuff :)
but im just thinking now that math itself should never be a goal, but it’s fine that people do play games of math for fun.
and i like how ai makes research, other work so fast and easy.
who knows & how could we possibly tell from here if all or any of the ai’s are on a good path for an actually better future than any one with different ai’s or even none at all.
i dunno too much about all of ai, but the pace of it definitely scares and worries me…🥲
it seems too frenzied. how could the best insights be had like this?