AI Writing | The Emperor with No Clothes
One of the bedrocks to quality content is deep research. Whatever level polymath you believe yourself to be, there’s plenty you don’t <yet> know.
As a hired gun in content, I’m often starting on projects within days of learning that the subject even draws interest. I figure that if someone wants to read about it and my client doesn’t want to do the writing, it’s time to pop a modafinil and make some magic happen.
Step One be like:
- you have a subject in mind
- you’d like a definition of that subject written
And we iterate from there.
But now there’s a fresh new artificial intelligence spigot. ALL HAIL.
You just rub the knob, mumble a quick incantation and…
On the surface, this seems so straightforward. Until you press a ‘write for me’ button and the AI sets about making something up that’s useless. Or, even worse, dangerously wrong.
There’s probably no better example of the need for quality research BEFORE you fire up an AI writer. Trust me, I want to see AI become powerful enough to fact-check itself but, like most modern news networks, it’s just as likely to tell you that Bono is serving 35 years for murdering The Edge with a guitar pick.
I like research, honestly. If I can suppress my natural inclination to try writing too soon, the whole process of learning a subject is quite fascinating.
But I don’t think most buyers of these tools give a hoot about creating useful content. They seem convinced that wildly inaccurate gibberish leads to revenue. I don’t think that’s true, though, as evidenced by how concerned they are about the cost of one of these tools.
The technology behind GPT-3 today puts it at just about the level of a random content generator. It’s like trying to drive a car with tires of four different sizes down a busy interstate.
Over the scope of my long copywriting career, I’d estimate that somewhere around 80% of my project time is on research. The other 20% is made up of writing drafts, finals, measuring results and updating.
So, if we decide that an AI writer can help me get draft words in place quicker than the traditional process, we are only really whittling at a component no greater than 10% of my job.
I’m certainly down for shaving off 3-5% of the overall effort but that’s not exciting enough to shove recommendations to go buy Jarvis or Scalenut at my list of email subscribers.
I’ve gotten some of these emails and may, one day, take the time to compile a list of the ridiculous subject lines and promises for your reading pleasure. If you’d love to know who has no soul AND no actual copywriting work to keep themselves busy, let’s review a few of those priceless ditties together!
That’s my rant about AI for today. I’m dying for it to prove me wrong.
