On AI writing – Dima Bolkunov’s Weblog
My bachelor’s thesis was about figuring out the writer based mostly on a brief paragraph. The concept was to check phrase utilization statistics in opposition to different authors’ texts. Surprisingly, even this straightforward concept labored fairly effectively. Readers use greater than phrase rely: patterns, sentence construction, methods of telling a narrative. It makes it even simpler to really feel the fashion.
AI averages fashion, so it’s arduous to see the writer behind it (even with numerous immediate tips). Empathy turns into more durable with out the fashion. When you realize the author personally, it turns into more durable to see the textual content from the writer’s perspective if you realize it’s not her fashion. But even you probably have by no means met an individual, non-AI-written textual content tells you one thing about this individual and their context. Ironically, with AI, context is being misplaced.
Beyond fashion, AI-generated textual content loses precision. Even given an in depth immediate, AI must make assumptions whereas writing to sew a textual content collectively. It’s the identical when you have to write a textual content for one more individual. This typically means selecting extra ambiguous, much less exact language to keep away from logical errors. For the author, it may be arduous to identify, as they’ve a transparent concept of their thoughts, which provides an ambiguous phrase a precise that means.
Another supply of dropping precision is spending much less effort and time difficult our personal concepts. Putting issues on paper normally helps us see inconsistencies and gaps in our judgement, that are simple to overlook after we are simply desirous about them. AI writing takes away this chance, as you soar on to a end result.
The lack of precision and elegance impacts studying velocity. You want to change ambiguous phrases for clearer ones whereas studying, and you’ve got much less details about the context to take action, as it’s misplaced because of the lack of fashion. Personally, I really feel that I’m at the very least a 3rd slower studying AI-generated textual content, even when the size is similar. Beyond this, most fashions are extraordinarily verbose, which makes complete studying time a lot increased. You don’t want many readers for the author’s time financial savings to evaporate on the readers’ aspect. This is even earlier than you’re taking second-order results under consideration: extra questions, extra iterations (if it’s a collaborative textual content), and fewer belief between readers and the author.
