A lot of people have AI automation all wrong.
They expect full automation right out of the gate – stellar copywriting or content writing, without fine-tuning the model, providing sample copy, instructions, or anything.
So of course it won’t get great results…
Currently, ChatGPT is best for augmentation, not full-scale automation.
Yet when it doesn’t meet unrealistic expectations…and despite the fact that AI has appeared out of nowhere to and can suddenly generate infinite amounts of text that is better than many humans…the attitude has shifted from one of amazement to “oh, it sucks now so it will always suck.”
But AI hasn’t even gotten started yet.
This is literally just the first inning…and, importantly, it can already fully automate many tasks better than humans, at a much lower cost, and much, much, much faster.
If you are not prompting it correctly…and if you’re using ChatGPT instead of the OpenAI API or the Playground Assistants…and if you’re not feeding it sample copy, then of course what comes out is going to be weird.
(I actually think OpenAI did that on purpose, but that’s a different topic.)
But you don’t just take the words it gives you and then stop.
You can refine the output repeatedly, using things like samples. And:
- Tell it what readability level you want to use (e.g., 8th grade reading level)
- Tell it what phraseology to avoid and what phrases to use
- Tell it to be more concise, more formal, less formal, use third-person, etc.
You can give it a lot of different instructions.
And you can provide it with copy to rip off…er, “sample.”
When you start doing things like this, the copy gets a lot better.
The unrealistic, unfounded expectation is that ChatGPT should be able to magically create better copy than humans, while leaving humans completely out of the loop.
But with a lot of copywriting, it’s about a spectrum of augmentation to full automation.
You just follow the trend line:
I used GPT-3, and then I used 3.5, and then I used 4.
And it keeps getting better and better and automating more and more.
And Sam Altman recently said that the next model is going to be way better.
So, since ChatGPT and Midjourney and Runway ML literally came out a year ago, I think we should stop acting like their automation impact has leveled off.
We don’t know that.
I think most people should listen to the only people who are experts: the people developing actually developing AI.
And these people, like Sam Altman and Kai Fu Lee, literally say that we’re all going to be out of work.
If they’re right, there is a possibility that today’s white-collar jobs will go the way of the farmer.
99% of them will disappear and be replaced by machines.
And what will be left?
“Gold-collar workers”?
I don’t know, but I’d rather be safe than sorry.
P.S.
An example of how to use AI to augment your writing (rather than attempt “full automation” without proper preparation):
- I spoke this post into a Google Doc, which produced a big long incomprehensible text stream
- I gave that long text stream to a Custom GPT that I built specifically for cleaning up transcriptions
- Once it was tidied up, I copyedited the transcript
- Then I gave the transcript to another Custom GPT I made specifically for generating Midjourney images
The result:
- A complete direct response copy-style post in a fraction the time it would’ve taken me to write it manually
- A lighter cognitive load (very important for writers)
- 40 Midjourney images to choose from (no need to search for or pay for stock photos)
Along with other white collar job roles, I expect AI to drastically change the way writers work, their role in the company, and the value they add.
While many may not like it, there’s not much we can do about it.
The only solution is to adapt…and because AI is moving at such a breakneck pace, this means adapting as soon as possible.
