One of the most frustrating aspects about generative AI is that it’s presented as a powerful editing tool by those who understand tech, not the practice of editing.
Yes, there’s the nuts-and-bolts stuff like dangling modifiers, tense agreement, accidental repetition, etc. But that’s not editing—that’s proofreading.
Real editing requires understanding the theme of the work, and then helping the writer maximize the emotional resonance of the piece by improving the music of the prose in service to that theme. These are human decisions—ones a predictive aggregate cannot do, especially when the writer’s approach to the theme is novel.
It’s mistaking writing for data transfer, which is never its effective element.
One of my highlights in undergrad was a writing seminar where we banged out a 3000 word essay every week. Early that semester the prof paired me up with this guy to do edit on each other’s work. Even though this required an even earlier deadline we liked it so much we kept it up for the rest of the semester.
I think your linguistic distinction between editing and *editing* was the source of my underwhelm. AI did a fine job sanding off the rough edges of my writing (for good or ill) but that was nothing like that semester working with Finn.
Given the rapid pace of advancement in AI I’m not gonna bet against this capability in the future (though what an impoverished experience compared to teaming up with another person struggling through their essays!)....but also one day my full self driving car will come.
Very interesting experiment!
One of the most frustrating aspects about generative AI is that it’s presented as a powerful editing tool by those who understand tech, not the practice of editing.
Yes, there’s the nuts-and-bolts stuff like dangling modifiers, tense agreement, accidental repetition, etc. But that’s not editing—that’s proofreading.
Real editing requires understanding the theme of the work, and then helping the writer maximize the emotional resonance of the piece by improving the music of the prose in service to that theme. These are human decisions—ones a predictive aggregate cannot do, especially when the writer’s approach to the theme is novel.
It’s mistaking writing for data transfer, which is never its effective element.
Ohh thanks for this comment!
One of my highlights in undergrad was a writing seminar where we banged out a 3000 word essay every week. Early that semester the prof paired me up with this guy to do edit on each other’s work. Even though this required an even earlier deadline we liked it so much we kept it up for the rest of the semester.
I think your linguistic distinction between editing and *editing* was the source of my underwhelm. AI did a fine job sanding off the rough edges of my writing (for good or ill) but that was nothing like that semester working with Finn.
Given the rapid pace of advancement in AI I’m not gonna bet against this capability in the future (though what an impoverished experience compared to teaming up with another person struggling through their essays!)....but also one day my full self driving car will come.