I was thinking the other day about the question "What is AI *actually* going to be good for?" It's hard because the answer isn't simple like blockchain ("nothing") -- in spite of the hype, there's SOMETHING there, but it's hard to pin down what.
It’s going to be significant for any non-literary writing, I think. Reports, e-commerce listings, corporate internal comms. Formulaic stuff where there’s a ton of stuff to train it on, then it works from new inputs. It’ll still need an edit, but it’ll make a three-author job an AI+one-editor job.
Will it be as good at those things as a human professional? No. But the mass produced sofa I’m lying on isn’t as comfortable or attractive as a handcrafted artisan piece and hey, it was cheap and does the job.
I think it depends strongly on how hard it is to check the text for errors. The problem with AI isn't the quality of its prose, that may improve. The problem is its tendency to make things up.
If checking for errors is time-consuming, then the AI doesn't actually save time. The margin is pretty thin, because producing basic text is not actually very hard so the savings isn't great.
The hard part of generating a business-style report, for example, isn't generating the text (which is often boilerplate anyway) but *gathering the data*. But that's exactly what you can't trust the AI to do.
