Becoming an AI Editor: Part I

So you’ve taken the generative AI plunge as an AI editor. You’ve written a detailed prompt, maybe even with sources you provided, or at least vetted after asking the large language model (LLM) to do some research. Now you want to do the right thing and put the human in the loop for the written output. But where do you start to do this efficiently and effectively?

You have three top areas of prioritization. Ensure that it’s good, that it’s right, and that if someone asks questions, you can answer them.

Let’s do these in reverse order.

Before you dive in to deeper editing, read the whole thing from start to finish. In order to be able to answer questions about the text, first ask yourself, based on what you know about the subject matter, does what’s written here make sense to you?

As editors of human-written content (weird thing to say, but here we are), this is one of the fundamental skills we’ve needed to master. Just because we have AI models now writing the text, it’s not time to give that up.

So what does “does it make sense?” mean?

First, it doesn’t mean necessarily fact checking every statement (though it may, and we’ll get to that in another post).

Primarily it means, given the whole of the piece, is there anything that seems wrong just based on the content of the piece itself? For example:

Are there any internal contradictions?

This is pretty self-explanatory, but still critical to watch out for. And it doesn’t have to be two contradictory facts. For example, maybe you’re talking about an established concept, but the LLM has decided to frame it as a new, ground-breaking idea. Watch for inconsistencies within the piece, and just using your own common sense.

Is the introduction of concepts done in a logical order?

As a simple example, take acronyms. We learn as humans to spell out an acronym fully on first use, so the reader has context for later abbreviations. Make sure foundational knowledge comes before shortcuts or anything that builds on that knowledge.

Is there anything that just doesn’t pass the sniff test?

A common problem for LLMs is hyperbole: something is “the best” or “the first” or “the only” or just “revolutionary.” Based on what you know, is it worth being as excited as the LLM is? Or are there any statements that just seem weird?

So just yet, don’t pull out your research library, but as a first step, read the piece and ask yourself: does this seem right? The answers may all be “yes,” or you may have some re-prompting/re-writing ahead of you. But either way, you’ll be prepared to answer questions about the content when someone asks you about it.

Next up, we’ll address when to pull out that research library. Stay tuned.