6 Great Points for Humans vs. Claude

I’ve got a couple of manuscripts just about finished and ready to be edit. In one case, the client wants to have the AI program Claude analyze the 82,100-word document before moving on and hiring a human editor.

His rationale comes from his time as an orthopedist. Many years ago, he was introduced to arthroscopic surgery, which had been around since 1919 but had made significant technological advances in the 1970s and ‘80s, when he began his practice. My client realized, perhaps before most, that this was the wave of the future. Therefore, he sees AI the same way.

So, he took my ghostwritten manuscript and sent to through ChatGPT and Claude. Then he told me about it, spending most of the time detailing how these two large language models (LLMs) fell short. 

“Keep an open mind,” he instructed me.

I did, and while most was nonsense, there were a couple of valid points. This told me that he might be right about AI being the future, but the future is not today.

Unbeknownst to him, I had asked several editors for opinions about chapter 1. One gave me a detailed analysis that touched on some broad areas that could or should be addressed. These included: ensuring the voices of the two main characters was consistent; that their insights, opinions, and outlooks of events they both witnessed and experienced are presented less chronological and more by theme; and striking a better balance between telling and reflecting.

In other words, he was doing exactly what my client was asking Claude to do. So, I asked this editor what I should tell my client to convince him that humans are superior to AI. 

His response:

“The most cogent point is that, because of the highly imperfect nature of current LLMs, an AI-based writer like ChatGPT or Claude is arguably even more in need of a human editor reading after it, to make sure it hasn’t introduced a lot of nonsense into the manuscript! (A lot of my job these days is fixing what AI writers do.) So he may just be introducing an unnecessary step into a process that probably should involve an editor eventually, one way or another.”

I emailed my client this answer, but still wanted to let AI do its thing. So, I added the couple of points, but my contention remains the same: Humans are better despite being, well, human.

Here are some pros and cons of using a human versus using AI. You will see that for every human con, there is an AI one. 

Pro: Humans can understand and recognize nuance and other subtleties. Being human means possessing the capability to recognize emotional subtext, symbolism, tone, voice, and character psychology. In other words, a human can understand what the story and book mean, not just what it says.

In my client’s story, there’s a great deal of character psychology as one brother falls in with gangs while another steers clear. AI can’t get into the character’s head and divine the motivations like a human editor can.

Con: Humans are slower and more expensive. It’s true that it took Claude less than an hour to analyze the entire manuscript—but Claude had to do it more than once because it took a human (my client) to give Claude the right direction and instruction to do what he wanted.

It’s also true that human editors cost about $5,000; Claude costs about $20 a month.

However… there remains something called value. A human with substantial editing experience is more than worth the investment. Plus, as the editor above said, he’s already cleaning up what AI is doing, so why not save the $20 a month and go right to the human solution?

Pro: Humans possess strong developmental insights. A human can help with big-picture issues such as story structure, pacing, character arcs, narrative tension, and ensuring the theme(s) is/are consistent throughout.

A human can recognize inconsistencies in how characters behave and interact. They can ask insightful questions such as “How does the protagonist change as the result of this event?” and “Why is the antagonist behaving this way when just five pages earlier, he acted differently?”

One of my favorite questions I ask whenever a client makes a suggestion that I find detracts from the story is, “How does this help move things along?” AI can’t yet do this and instead tends to give broad and formulaic feedback that might not best serve the manuscript.

Con: Humans have biases. Each editor has his or her own tastes. One editor loves something that another dislikes. A program like Claude can be far more neutral.

Except for one thing: LLMs get all their data from humans. That inherently means the biases are built right in. I’ve seen and read reports of AI programs refusing to admit they’re wrong when the facts show they are. 

It is important to find a human editor that will put as many biases aside and give the manuscript the neutral third-party set of eyes it needs.

Pro: Humans know the market. If there are trends in publishing, if there are sudden monumental changes, humans will know first and react appropriately and quicker than Claude or any other AI program.

A human knows what the expectations are with any genre. And if one human knows he or she doesn’t know, he or she can ask another human who will know. This recently happened to me. I was working on a different project in which the client wantsedto start the book at a certain point and then come back to that point using the same words as at the book’s beginning.

I didn’t know if that was acceptable in the publishing world, so I emailed three publishers, and they all said it could work, “as long as the timeline shift is clear and easy for the reader to understand, so they’re not left feeling confused and wondering what’s happening,” cautioned Christine Kloser of Sheshat Press.

Con: Human feedback is limited by cost. While my client could ask Claude to go through the manuscript an unlimited number of times, humans want to be paid for the time. Therefore, many editors limit the number of times they go through the manuscript. If more is needed, the client pays them.

My response is that humans can be efficient if you find the right one.

The bottom line is this: AI is here to stay, but humans would be wise to know when to rely on it and when a human is still the better way.

Feel free to read and check out my other posts related to ghostwriting. Go to https://leebarnathan.com/blog/