Pssst… this piece on GPT-3 for copywriting was sent out to my newsletter. Once a week (ish), I send an email on copywriting, persuasion, business, or… whatever’s compelling and useful to me at the time.
Curious? Sign up!
It’s the sound of thousands of copywriters collectively crapping their pants the moment GPT-3 dropped. That, or the dripping of sweat as many writers (yet again) start wondering: how far away are we from a robot stealing my job?
Even the notion of an “AI that can write copy” tends to put a bee in the bonnet of writers everywhere:
“AI can never do the creative work of a writer! It can never understand an audience! Anyone who would use a robot to write for them is an evil turd gremlin!”
But whether you’re a writer or you just hire them, you need to be aware of GPT-3, because it represents…
- A potentially seismic shift in how written communications come together,
- A HUGE potential ally for all of us, and
- A fairly considerable threat to some of us.
What is “GPT-3” anyway?
The oversimplified version: ‘Generative pertained transformer-3’ is a language-generating AI:
“GPT-3 can execute an amazing bandwidth of natural language processing tasks, even without requiring fine-tuning for a specific task. It is capable of performing machine translation, question-answering, reading conceptual tasks, scripting poems and elementary maths.”
That’s doesn’t SOUND sexy, but you should take a look at some of the incredible things GPT-3 can do here: https://mobile.twitter.com/xuenay/status/1283312640199196673
Also impressive is that you can train the AI on your own writing and it can learn how YOU communicate in terms of voice, tone, etc
Is it perfect at writing copy? Heck no.
Some of the applications I’ve seen show GPT-3 doing things like composing full emails from bullet points. It’s impressive until you scrutinize the copy and see three statements beginning with “I’m not sure…” in a row. But then—those are easily and quickly edited by a human when the bulk of the writing has been done.
Are writers’ jobs immediately canceled? Again, heck no.
One of the major weaknesses of GPT-3 is that it cannot understand the words it writes, so it makes silly mistakes.
Even in this early format, the power and potential is pretty startling, and to ignore the impact of GPT-3 on copywriting means you’ve got your head shoved firmly in the sand (or other undesirable places.)
Here’s what I think GTP-3 means for copywriting:
+ If you’re making your living writing SEO-oriented blog posts to target long-tail phrases where creativity and branding are not critical, your days are legitimately numbered.
GPT-3 can reliably put together content on topics like these that will only need minor editing from a human to be ready to be pushed live. This could legitimately eat the types of simple jobs that roll through Upwork/Fiverr alive, depending on what access to GPT-3 winds up costing.
Regardless, the role of “editor” just got more valuable
+ Similar to long-tail blog posts, the same may go for (some) product descriptions, location-based landing pages, etc. — anything that is fairly programmatic, usually done at scale, and doesn’t need to be amazing. The SEO industry as a whole is going to bite into this thing like a leech and bleed it for all they can.
We’re a ways off from AI that can single-handedly write compelling sales pages, brand a business, or dream up creative ways to sell pain points and benefits—though an AI that can do so much on the permutations side can sometimes sound like it.
On the positive side of things…
+ I believe we’re about to see the advent of some INCREDIBLE writing tools!
Tools that can…
- Analyze thousands of reviews and help isolate the intel we use to write.
- Generate solid first drafts of formulaic content to punch up and expand on, dramatically speeding up the writing process.
- Identify patterns across disparate data (think: reviews and chat logs) that no human could account for given bias and limited time to analyze.
- Translate complicated sentences into plain-speak (already happening!)
- Take core ideas and spit out loads of variations (e.g. huge numbers of headlines, ad variants, etc.) — in fact, we’re already there, with platforms that can produce HUGE numbers of targeted ad variants for testing.
I also believe…
+ Skills like editing, researching, and data entry may be worthy pivots to add to your toolkit, but more importantly…
+ Strategy has never been more critical to learn. You want to be the one helping steer, correct, and feed the machine, not the one losing production work to it.
What’s on the horizon (sometimes near, sometimes far):
+ In the not-too-distant future, you’ll be able to write what you want a piece of code to do and have AI write that code for you.
+ You’ll be able to describe a basic design/layout you want for your site and have AI build it for you.
+ College essays are going to be tougher to grade.
+ Fake reviews are about to get a lot harder to spot, as is fake news. I’m legitimately concerned about this one: in a world where so few of us critically evaluate what we consume, the power to create infinite fake news is scary.
+ Text-to-voice tools (like Lyrebird), plus deep fake tools, plus NLP tools like GPT-3 could lead to some REALLY interesting outputs, like characters in video games who can have real conversations with you related to ongoing/dynamic quests, spam calls that sound real but have been auto-generated using whatever records on you the robot can find, or even contrived podcast episodes where no human was actually present!
Seriously, the future is terrifying and exciting all at once!
To sum up:
I don’t think copywriters’ jobs are about to disappear overnight. If anything, I think we’re about to get some amazing sidekicks in our corner to make our writing all the more powerful!
But with great power comes great responsibility, and I’m not sure the world is prepared for how to deal with the tech we’re creating.
Great read; thanks. Definitely important not to write yourself into a corner (ba-dum tsh) as the landscape evolves.
The technology is developing rapidly, but it is still very far. AI is more HYPE than reality at the moment. Sure, simple things can be done, but writing meaningful text that can match the quality of experienced writing is far in the future. Computing power is growing, but this wont help much into AI learning process. Just remember, first sentences of every child starts at the end of the first year.