Artisanal Writing

Prediction: Generative AI platforms built using large language models (or some development of such technologies) will be used for a majority of transactional writing tasks and many cases of creative writing tasks (with humans in the loop). A category of pure human writing will come into being, marked as distinct. Call it **artisanal writing** if you like, similar to the artisanal movement around many products that were all mechanized.

The meaning of this human-crafted writing will change over time and it will be used in certain contexts where before it might have been the default for all contexts. It may even be the case that special styles will develop, proclivities for certain markers of human-ness in writing. These may not be purely stylistic and may accelerate the adoption of novel or insider lingo in order to keep one step ahead of the AI training. (We already see the first wave of attempts to thwart Bing's gpt-4 engine by injecting specific commands along with the real content of things.)

Every use of generative AI in writing platforms seems to be some sort of writing assistant that does one of two things. It either generates a bunch of boilerplate or templatized content or it gives you some (largely bs-y) assessment of how well your writing does or doesn't fit into a set of boilerplate criteria. This is the way of the LLM, trending towards the average and conformity. It's like the orthodontification of writing. Everyone's writing is straightened to the point that there is no longer anything distinct or remarkable or interesting. All novelty is algorithmically constrained, the regurgitation of the past with the regularity of randomness.

Maybe language models will get better.

Or maybe we will need to niche ourselves down and outrun the stylistic flattenifiers.