If AI has accomplished one thing, it’s been prompting everyone to expect the worst,. We’ve jumped on the “ AI will take jobs away” bandwagon. We are alarmists. But if you look back at some of the alarmist things we have said in the past and how the reality has been different, you would think better sense would prevail.
Content writers are no different, antsy that the bots are coming to write great content on short notice for almost free.
Is it time for AI to start writing your content? Should you fire your content team? Is it time for writers to search for other jobs?
The answers are no, no, and no, respectively.
But I’m human, too, and I got nervous. I was reading articles about AI content writing and wanted to try one of the bots to see how it is. I scanned the internet and found one that had great reviews in top publications. I was not hoping for a finished article anyway but merely a good start. I was disappointed. I entered logical words in meaningful order. I got back a word salad. Try one and see what you get.
It’s not just me, a lot of experts think AI content writing is overrated, too. I’ll concede that we have made considerable progress in this area, but even now, AI tools are not advanced enough to snatch away the job of writing impactful content from us humans. Before I explain my logic, let’s look at how AI content works and what AI writers can do.
How do AI content writers work?
A lot of what is happening in the AI content creation world is based on Natural Language Generation or NLG. Natural language generation is a software process that primarily deals with converting data into descriptive text. The idea behind this is to generate human-readable content from data points in a scalable fashion.
Currently, AI content writers that are commercially available take some input in terms of data or even a heading and string together sentences it believes will be useful based on the inputs. NLG is still in its very early stages and has seen a few commercial applications already.
What is GPT3?
GPT3 has gotten everyone quite excited. It is the flavor of the month in AI enthusiast circles. GPT3 is a language model that has been developed by OpenAI. GPT3 is the most complex and the most powerful language model created, with a massive 175 billion variables to work with.
GPT3 also has been trained on a great deal of content– nearly all content on the internet, which comes up to 45 terabytes of data. All this training has enabled GPT3 to generate near-human-like language. A lot of what GPT3 does is, it learns from the data based on what it receivs as input and then creates output in the desired format. For example, you can feed in an Excel file and generate a report from it, even!
What can GPT3 do?
Currently, OpenAI has released an API that organizations can use however they want. People have used GPT3 in a variety of ways, such as:
- A chat tool that responds as if a human would. (Like an AI friend?)
- Summarizing text: A couple of tools created summarises a block of text into a few short paragraphs.
- Translations: Translate from one language to another. An English to French translator has been created by an organization based on the GPT3 API.
You can find many more such examples. But reading the other experiences of users like myself, all eager to experiment with a new toy, we see one clear use – AI content writing.
If you use the API to input a headline and some basic data points, it can generate whole paragraphs and even full blogs, like this one here.
What can AI writers do?
Going by the current areas in which AI writers already exist and the various GPT3 use cases, we can see that AI has a role to play in content generation. Currently, many leading news organizations rely on AI writers to generate quick and short data-based pieces with no editorial input, things like weather reports, financial reports, or sports summaries. Basically, AI can fill in wherever data needs to be presented in a determined format. This tool frees up time for journalists to complement these stories with more in-depth stories.
AI writers have proven quite proficient in codifying data and information from across the internet to expand on sections, write ad copy, generate content ideas and summaries, and much more. Note this, however. The AI tool isn’t writing jack. It’s just rewriting what already exists in (hopefully) grammatically correct language.
What AI writers can’t do… that humans can
So I’ve granted some space to the AI content writer crowd. The bots can do a little. But not enough. Here’s what we can do that they can’t do:
Write about emotions and experiences
Writing is a lot about emotions, both in the writer and in the reader. AI does not have the capability of talking about emotions in the way that humans can. No matter what the movies and sci fi novelists tell us, no robot feels human emotions. Most of the articles/blogs you create will also have a certain mood about them, which AI writers would find impossible to imitate.
Plus, AI does not understand the difference between lying and telling truth. AI writes look at what’s written, assumes the data is true, and goes with it. So a lot of bias, half-truths, and full truths creep in.
Writing also involves going back to our memories and experiences to draw inferences and references. AI writers go by the reports of many people, and hence, it is not easy to convey authentic, unique experiences.
Talk about something new
AI writers look at everything that has been written about a topic and link together words to make sentences. So everything an AI writer does depends on what has already been written. AI writers can’t create anything new from the data. Conveying new ideas and new ways of looking at things is what a lot of today’s writing is about, and AI writers can’t imitate this yet. I don’t think they ever will.
Write content that doesn’t suck
I tried saying “falls short of standard” here instead of “suck,” but it did not convey my real feelings about the content AI writers created for me, so let’s go with sucks. AI’s content just does not hold up well against scrutiny. It’s usually a word salad with a string of words tossed together that does not offer readers new information, valuable data, or even a pleasant reading experience. It’s almost like the order you put the words in matters.
You can even find a shocking amount of bias and hateful language if you venture deep enough into machine-written text. This problem has been one of the major flaws in the AI design. Beyond this, the AI is prone to making up things, too. Human writers can avoid doing that.
Bringing it all together
AI writers have come a long way, especially riding on the GPT3 hype. But the fact remains that it is what it is – a hype. Not just my words. This is what Sam Altmann, CEO of OpenAI said: “The GPT-3 hype is way too much….AI is going to change the world, but GPT-3 is just a very early glimpse.”
It’s not entirely useless. AI can solve a lot of issues in the content domain, especially in keyword research and SEO. It can also do a lot of the work that involves converting sets of data into descriptive statements based on specific parameters. These features are great and meet certain content creation needs.
AI content writers are definitely going to play a role, but that role will only supplement human-voice content. It is not going to replace human-generated content, at least in the near future. It remains to be seen what else AI can achieve and how we can leverage its power.
Until something big changes, though, our writing jobs are not going anywhere.