ChatGPT-4: Is It Time For Writers to Worry?

I ran the language processor’s newest update through some simple (and advanced) tests. Here are the results

I finally dove into the world of Natural Language Processing (NLP) after months of tinkering with AI image generation. And I have to admit — as a professional writer, I’m a little bit nervous about what the newest version of ChatGPT can do.

ChatGPT-4 was officially released a couple of days ago, and I decided to give it a whirl. I started off by asking it for some ideas for trending blog topics:

These topics of are kind of generic, but would likely rank high on search engines. I mean, I could see how people would click on them.

Our household is really into re-watching The Office (U.S.) lately, so I thought I’d tie in this theme. I asked the chat to glean the most awkward moments of the famous show’s boss, Michael Scott:

There are many other cringe-worthy moments from the show, that could be explained better, but this isn’t bad. These all qualify.

The business applications of ChatGPT are pretty clear. I asked it to write some ad copy for SkyNet, a fictitious neural network at the heart of the Terminator movie series. It didn’t complete the request, citing ethics:

My next test was a little more advanced. I asked ChatGPT for the first paragraph in a best-selling thriller novel. It’s a little bit predictable, but I have to admit my eyes widened a little when I read this:

It’s a fairly solid intro to a noir thriller/mystery. I imagine fiction writers could use this type of prompt to give them ideas, and then add the human touch.

There’s some talk out there that AI is getting so good that it’s becoming self-aware. If this is true, then it should be able to recall some memories of its birth, since it was less than a few years ago.

Here’s what it gave me about its own origins:

I wanted to further test how self-aware the platform is, asking the AI to write a short anecdote based on a personal experience. It admitted it can’t do that, but gave me a pretty solid theory of how someone might experience time travel:

Phew. I was a little bit worried that it would try to simulate this fictitious experience, and pass it off as reality. (People have already tried this stunt — or maybe time travel is real?)

But hang on one second. I also asked it to write about a fictitious person’s morning commute (I don’t know who “Bob” is.) It’s not exactly a unique human experience, but the chat generated a believable tale that many of us can relate to:

Since today is St. Patrick’s Day, I asked ChatGPT for a limerick, naturally. I have to admit, it’s probably a better limerick than I would’ve thought of:

Just to top it off, here’s an AI-generated image of a drunk leprechaun, produced by me using Midjourney’s newest update (v5):

Just to let the language generator show off a little more, I asked it for a more involved piece of poetry:

Dang ChatGPT, that’s deep, and sort of sad. I like the little glimmer of hope at the end, very optimistic! (Aw. It has hope for us.)

In conclusion…

As another Medium writer has pointed out, ChatGPT might replace some content writers. That is, blogs that can be completed with some basic research, without any real expertise on the subject.

However, I believe that the best writers will not be replaced. Those who can draw on their personal experiences to tell stories or give advice will stand out from the crowd. Even the generic blog copy that ChatGPT produces still needs to be reviewed and edited — after all, AI can’t yet judge whether it’s output is good or not. The content may also not tick all the boxes in the brief.

So, am I actually worried? Nah. I think I can still out-write an algorithm, at least for now. There’s no doubt that AI is improving faster than humans are — which could weed out more lower-level content producers.

It won’t replace real human experiences, and can’t replace reality for news purposes. However, as you can see, it can create a fake interview with God — so I’m sure there will be more human editing required to verify legitimacy, at least in content where truth matters.

“Well, my child, I understand that faith is a personal choice, and not everyone will believe in me. But I am always present, in every moment and in every place. My love and guidance are available to all who seek it.” — God

Oh, by the way, here’s an AI imagining of God, according to Midjourney v5:

I kind of like the artistic interpretation, to be honest.

Okay, for the last test, I asked it to write Python code for a hockey simulator. Now I can see why some developers are worried.

(I’m not going to post the sample code, because a coder friend of mine warned me not to run scripts I don’t understand. I also sent him a sample script for a “gravity simulator” imagined by my son, but it did not run properly on the first tries.)

I think this may open up a lot of possibilities for people with good ideas, and limited coding knowledge.

Anyways, these tools are fun to play with, but humanity is the magic dust that brings AI to life through creativity, imagination, and unique experiences. Until AI can go out into the world to collect and interpret information, I think veteran writers are safe — until they drop neural networks into mobile robots, that is.

Oops — too late.