Ever since Chat GPT came out at the end of 2022 I’ve seen predictions about content. Some good and some bad. But the general idea is that ‘the cost of content is going to $0”. This is hot-take based on the hype-cycle around a new technology. This prediction doesn’t follow any historical cases, and it assumes that creating content is primarily writing – which it isn’t.
There are some highly relevant historical examples, creating content is more than just writing, and there are fundamental flaws in Chat GPT that humans need to compensate.
Table of contents
Historical Examples
Virtually every time a new technology comes out people panic that the sky is falling. People aren’t comfortable with uncertainty and assume that if you can’t see where you’re going you’re 100% going off a cliff. This just isn’t true.
The future is always uncertain and while we can’t predict exactly what will happen with Chat GPT we can look back at another piece of software that eliminated 98% of a job.
The History of Spreadsheets
Yes – I get to talk about one of my favorite topics: spreadsheets. No not the cool sleek Google Sheets that I use daily but the old kind. The kind that teams of accounting clerks and bookkeepers would pour over and manually adjust each cell.
Prior to 1980 companies employed teams of bookkeeper and accounting clerks and they would ask them a big business questions. Ex
- What if we increase production by 5%?
- What if we give everyone a 7% raise this year?
- What if we lower the cost of one of our ingredients by $1.00/unit?
Those accounting clerks would collect the data, run the data through the spreadsheet manually and then present the results to the executive team.
This manual data processing was incredibly time intensive. One little mistake meant a whole day of erasing and filling the same boxes out again. It used to take weeks or months to produce results with a manual spreadsheet.
In 1979 VisiCalc was released for the Apple II computer and it was the first modern spreadsheet software. Nowadays with formulas and processing power we can do an entire weeks worth of data processing in literally seconds.
So what do you think happened to all of those accounting clerks and bookkeepers?
The economics podcast Planet Money actually looked into these numbers via the US Census:
“[S]ince 1980, […] 400,000 bookkeeping and accounting clerk jobs have gone away. But 600,000 accounting jobs have been added.”
There was some job destruction. We lost a lot of bookkeepers but many of those bookkeepers retrained into accountants. Over a 10 year period there was 200,000 net job gain despite losing jobs early on.
What Happened with Spreadsheets Will Happen with Chat GPT
So what happened here?
There's a common fallacy that there's only a certain amount of work to be done. And if machines pick up some part of the labor then there will be less work for humans. This is called the Lump of Labour Fallacy. It is a fallacy because work isn't limited. There's too much work to do and not enough humans to do the work. Once machines learn to do this tedious work that frees up humans to new and different work.
With spreadsheets specifically, data was so valuable that businesses simply started asking more questions. Instead of running 4 big spreadsheet models you could run 400. Instead of just looking at reducing cost-of-goods by 5% they can now look at reducing cost-of-goods from 1% up to 20%. And then break it down further with accessory products, foot-in-the door products, and more.
They found value in more data for the same cost. The biggest constraint on the business was only willing to spend X% of revenue on analyzing and predicting data. Once digital spreadsheets proliferated they got more of what they wanted for the same price.
I predict the same will happen with content. Companies value content as a way to build trust with your audience, to introduce someone to your brand by answering their questions, and to help your existing customers solve their own problems. Companies will keep their marketing budget roughly the same and deploy additional content in a much faster way.
Creating Content Is More Than Writing
Do you think a pilot just flies? Or are there additional parts to the job like safety checks, navigating storms, taxiing, and takeoffs and landings, etc.
Do you think an electrician just connects wires? Or are there additional parts to the job like drilling holes, running wires, upgrading electrical boxes, etc.
Do you think a developer just writes code? Or do they also create database structures, upgrade old and outdated technology, and write automated tests to prevent future problems?
It’s easy to think that a writer just writes. Or a radiologist just looks at images.
Stop Training Radiologists
Geoff Hinton, an expert on artificial neural networks, said “Stop training radiologists now” in 2016. His argument is that AI systems that study radiology images so we won’t need radiologists in the future and we should stop training them now.
“Take any old problem where you have to predict something and you have a lot of data, and deep learning is probably going to make it better than existing techniques”
Geoff Hinton
Do you think he was right?
It’s easy to assume a job is primarily one or two tasks – like a radiologist primarily looks at images.
But in reality most jobs are a cluster of related and overlapping skills. We still need radiologists because they’re better at understanding a patients history, they’re better at collaborating with the rest of the hospital staff, and they can solve complex problems by working with other radiologists.
The same is true for marketers. Good marketers do more than write:
- They review your existing content so you avoid competing with yourself (aka keyword cannibalization)
- They keep an eye on your Google Search Console to catch errors early
- They organize and structure your website for users as well as for Google (aka cornerstone content)
- They know what proprietary data you have access to that your competitor's don't (aka maximizing your competitive advantage)
- They decide how to break a big topic into smaller pieces (aka keyword clustering)
- They log and track all of the keywords you’re using and take note of which posts are working and which aren’t (aka data analytics)
- As new content is created they review old content and add links to the new content (aka internal linking)
- They decide when and where to add graphics. And some magical unicorn writers even have graphic design skills to do it themselves. (aka graphic design)
- They ask your audience about their biggest problems, log the responses, and categorize them (content ideation)
Gary Vee thinks content ideation is where the new jobs will be:
I recently published the content creation process we use for eCommerceFuel and it’s 13 steps long. Three of those steps involve writing. The other ten are strategy, graphic design, internal linking, and more.
Writing a headline may only cost a couple minutes of your time but doing the job of the content manager is so much more comprehensive.
Overcoming Flaws in Chat GPT
Every tool has weaknesses. You don't want to use a hammer when you need a scalpel. The same is true with Chat GPT. It's an incredible word prediction machine. And if you need to predict the next possible word it's the perfect tool.
But most content creators aren't trying to predict words. We're trying to educate our readers. And we're trying to share an opinion with the world.
Chat GPT Hallucinates (AKA Chat GPT is a Bullshitter)
According to one AI, Elon Musk died in a car crash in 2018. He didn't.
According to Chat GPT George Washington invented the cotton gin. He didn't. That was Eli Whitney.
A laywer used Chat GPT and it cited bogus cases. Even after the lawyer asked Chat GPT if those were real cases.
Chat GPT writes things that are plausible. Truth is irrelevant. That means you have to spend a ton of time editing anything that Chat GPT writes.
If Chat GPT says “email marketing is the best marketing channel” is that true? Is that verifiable? Is it an opinion? You have to figure that out and make it clear in your article. If you read that line in a Chat GPT response you might spend 20 minutes looking for an article to prove a point that may not even be important.
Editing a Bullshitter Takes Time & Concentration
Chat GPT makes it easy to skip past the “blank page” problem. Chat GPT and other Large-language-models (LLMs) give you something that sounds plausible in seconds. But if you want to be careful with your reputation you will need to spend more time editing.
I've worked with experienced writers and I write myself. Writing an outline and that first draft is probably the hardest part. But that's a problem I know how to solve. I know how to timeblock 2-3 hours in the morning to get out the first rough draft.
Editing every single sentence that Chat GPT produces to make sure there aren't any errors is tedious and takes 100% of my cognitive attention. It's not a task I can do in a noisy coffee shop or a task I can do for 30 minutes before the next meeting starts.
Productivity experts agree there's a limited amount of time you can concentrate through the day. The common consensus is about 4-5 hours.
With a limit of 4-5 hours for all of my deep work per-day do you want to spend 2-3 hours of it editing a Chat GPT article? I don't. Other types of work like brainstorming and reviewing competitor headers can be fun and don't use that limited 4-5 hours of focus.
Final Thoughts: Content Is Never Going to Zero
Writing isn't going anywhere. We still have data nerds who play with spreadsheets. And we have a shortage of radiologists despite powerful AI's doing the most important aspect of their job.
Writing will still be here a year from now and 10 years from now. It's certainly going to change with new tools. And I expect most writers will use Chat GPT as a productivity tool.
I also expect Chat GPT to be paired with another AI that verifies claims. That would be hugely helpful to know that the content Chat GPT puts out is accurate.
Ultimately, you're paying someone to create content for you. Someone needs to come up with the strategy, analyze the strategy, and update the strategy as you go. They need to know enough about writing to get the writing done. But they also need to know what makes your brand unique. They need to know what proprietary data you have access to that another brand doesn't. That's how maximize your competitive advantage in writing. And that needs to be someone's job. And you're going to have to pay them for it. Writing is how they get the job done. But it isn't the job.