Provenance and AI

Recently, we had a client in the fine arts space, so the assignment required a kind of copywriting that ties to journalism in an interesting way. This is because when you have a collectible, you have a single piece of physical media, and when it has a backstory, that backstory may impact the value. The shorthand term for this is provenance, which is quickly defined as the history of ownership.

From an economic standpoint, provenance is hokum, fluff, spin, or worse. You bought a painting; the idea that your experience of it may be worth more or less based on who handled it before you seems like, at best, nonsense. And yet, provenance may color a surprising number of purchases in your life, because branding is like that.

The wine is local? Provenance. The band records live to tape with no editing? Provenance. The house was owned by a history of happy families? Provenance, or feng shui, or not on disputed lands or close to environmental concerns? That song you like is suddenly loved by people you don't? Whoops.

Now, it's clear that commodity purchases have little impact from provenance, but how many pure commodities does the average consumer actually buy now? Gasoline has detergents and octane and corn supplements, electricity comes from fossil fuels or sustainable ones, and so on. Making a commodity into a service is a time-honored way of making a profit, and provenance is all margin, after all.

Until, well, it isn't. 

If the provenance of the painting involved murder and theft (and yeah, the Nazis stole a lot of art), you'd have to be a pretty special kind of outlier audience to think that made it more valuable, rather than something that should just be donated in an act of reparation.

We don't really know why we attach values to objects, and maybe not even when we do. Provenance threatens to supersede that judgment, lay it bare and open for all to see, and change the way we value items in our world. If you are in the proper head space for provenance (i.e., not distracted by something more urgent, trusting of the source of information, looking to assign or impose a resale value), this is all welcome information. If you are not, it's chaff. 

So getting the tone of this requires effective storytelling, a connection to a shared humanity, and the ability to tell the user not just what they want to hear, but things they don't. 

Or, shorter... why would you value the output of any LLM or AI, when you know the provenance is so relentlessly mundane, for anything other than, well, commodity work?