the person getting the best results from AI art tools right now is someone who could art direct a feature film but can't draw
that profile didn't used to be valuable. now it might be the most valuable creative profile that exists (and there's no job title for it)
Machines replicate broad market taste perfectly fine. That's literally what they're trained on.
The scarcity is knowing when & how to pivot from that center, cuz everyone using the same tools toward the same training data produces the same output.
Applied taste is what creates the deviation.. that's where the actual distinctiveness (and value) lives.
@thenanyu AI aggregates revealed preferences at scale. Calling that taste collapses a distinction worth keeping.
Taste is what we can use to steer that aggregation toward a specific goal.
The moment you're optimizing for anything other than peak broad appeal, you need it.
narrowing the word with domain-specific distinctions does add helpful clarity...
but the conclusion it leads to ("taste has no meaning without a domain") is solving the wrong problem.
the issue isn't that taste is domain-specific. it's that there are two different things called taste & we're using one word for both.
the first is preference alignment: your aesthetic responses tracking reliably with what's good in a given domain.
that one IS domain-specific
you build it through exposure & it doesn't always transfer cleanly. great music taste doesn't port to architecture because the underlying pattern library is different.
the second is taste as a practice: the ability to take a preference (yours or someone else's), reverse-engineer what's producing it, form a hypothesis, make or evaluate something against that hypothesis, & read the result as on or off mark.
that one transfers.. not exactly because the preferences carry over but more that the analytical loop does.
the people who seem to have taste across multiple domains are ppl who learned to treat every aesthetic response as the beginning of an experiment rather than just having unusually broad exposure.
"good taste" & "well-defined taste" aren't the same thing.
Good taste is a relative measurement: your preferences aligning with (or ahead of) what a market or audience values. it can develop passively and exposure does much of the work.
Well-defined taste is a practice.. it's the ability to take a preference (yours or someone else's) reverse-engineer what's producing it, form a hypothesis, make something based on that hypothesis, & read the result as on or off mark.
not "i liked it" or "i didn't." on mark or off mark.
the distinction matters because one is a feeling & the other is "data" you can act on next time.
most creatives develop the first & assume they're developing the second.
but the difference rolls up the moment you have to execute precisely.. for a brief, for a client, for a collaborator whose taste you need to understand & serve.
that's when you realize "good taste" gives you a strong reaction but no efficient way forward
you can train the second one. but it requires treating every aesthetic response as the beginning of an experiment, not the end of one.
x is an irresolvability engine.
the content that performs best is structurally incapable of resolving (resolution kills the loop)
see: every "taste" debate happening right now
the word means four different things to four different ppl.. so the argument can't close, it can only grow
"voluntary departure plans" are the most honest euphemism in corporate history.
Every departure under restructuring is voluntary.. you either volunteer to leave or volunteer to do three people's jobs until you quit on your own