Skip to main content

luchot miscellany

A toy model of aesthetics with just two binary variables, 'classiness' and 'busyness'*:

Are these descriptions true? Well, they are incomplete, and are not definitions (i.e. one-to-one mappings), but yes. Are they helpful? As a start, absolutely.

Now, the labels on the left are vague and intuitive family resemblances; it is a fool's game to imagine they could ever be nailed down as monothetic definitions (the philosopher's ideal of neat, necessary and sufficient sets of attributes). We can still model usefully and harmlessly, even if the models can never be complete.**

But the critics and art academics I know spend far more time muddying the water: deconstructing our use of the problematic term "classy"; and who gets to say what 'simplicity' is anyway? They don't seem to want to explain things, even fuzzily.*** Or, maybe they do, but refuse to accept anything but a perfect final omniperspectival explanation (the like that can never be supplied), maybe to keep themselves in work.

Imagine if critics were conscientious enough to build a consistent hundred-variable, real-valued theory of art. Would it "solve" criticism? Never ever. Would it make the points of disagreement between interpretations more vivid? Would it force clarity in this, the most pompous and vacuous discourse? Yes.

But we will probably have to wait for AI art critics for that, to go with the excellent AI artists we have already.

There's no fixed criteria for these terms, you say? There's too much political context and social problematics involved for art to be tackled by statistical inference, you reckon? Well, machine learning is the automatic empirical discovery of non-necessary, non-sufficient attributes; it can and will cover the full range of the term's application and will do so by frequency, not political agenda.

The polythetic wall held up against philosophers and computers for a long time, sixty years at least. But it's time.

someone might object against me: "You... have nowhere said what the essence of... language is: what is common to all these activities, and what makes them into language or parts of language. So you let yourself off the very part of the investigation that once gave you yourself most headache, the part about the general form of propositions and of language."

And this is true. — Instead of producing something common to all that we call language, I am saying that these phenomena have no one thing in common which makes us use the same word for all, — but that they are related to one another in many different ways. And it is because of this relationship, or these relationships, that we call them all "language". I will try to explain this.

The result of this examination is: we see a complicated network of similarities overlapping and criss-crossing: sometimes overall similarities, sometimes similarities of detail. // I can think of no better expression to characterize these similarities than 'family resemblances'...
and then Nils Nilsson:
Some tasks cannot be defined well except by example; that is, we are able to specify input/output pairs but not a concise relationship between inputs and desired outputs... machines [are now] able to adjust their internal structure to produce correct outputs for a large number of sample inputs and thus suitably approximate the relationship implicit in the examples.

We can do this to anything we have at least proxy data for, which is, arguably, every thing that could matter. (If you count e.g. self-report of experience as reliable proxy data for consciousness.)

* A solid addition for a three-variable version would be "flat or shadowy" i.e. using clean planes or chiaroscuro. This would let us introduce Classical (simple classy flat) and Gothic (busy classy shadowy) and three others I can't be bothered looking up. Though this is a distinctively visual epithet, where the above should apply to all arts.

** What are the risks of building a model? Does a model obscure reality behind its necessarily limited representation? No; all the authors and users of models need, to avoid delusion and harm, is a little imagination and humility.

*** There are of course honourable exceptions.


An emotion killed by the internet: the sweet, terrible pain of hearing an incredible song on the radio and the announcer naming neither the band nor the song and no distinctive lyrics sticking out and no searchable databases in the world to trawl anyway, let alone a free machine to hum the melody to in a crazed offkey blart.


"Reification" is an insult in academic circles - Marxists and their descendants only use it for talking about delusions that make themselves real. (Adorno blames it for Nazism!) But every act of creativity and every act of love is a reification.


You immediately identified the above as a list of electronica artists - ones you can't dance to. (Some of you will have guessed that it was dark ambient and concréte stuff.)

And anyone can immediately see these are punk bands. Every genre has something similar: a limited vocabulary to denote themselves, an average political outlook, and a dress code.

Why? There is no real reason a particular sound should be played only by people dressed a certain way, nor only in groups named in a certain tradition. The answer is that musical genres are not just categories of music. They mark out categories of people (: tribes) - and tribes love inventing differences and clustering well within those differences, so as to not be mistaken for the hated other tribe.