Is artificial intelligence content writing possible?

How can you tell if you’re reading content created by artificial intelligence (AI)? Before I tell you, please identify the pictures with bridges to prove that you’re not a robot.

Apologies for the bad joke.

I thought that opening this article with my attempt at AI humour would confirm my true identity as a human. I mean, when ChatGPT came online in November 2022, over a million people signed up in a week because it conducted credible conversations.

I was wrong. I asked ChatGPT for the best joke about AI and got this:

I’m sure the jokes will get better, but as AI generated text becomes banned from schools, labelled as spam by Google, and delivered as personalised Buzzfeed quizzes for you – there’s no doubt that AI will be playing an increasing role in all of our lives.

So, how do we play nice with AI? Will AI make our jobs redundant? What are the ethical challenges of AI? And, dare we ask, is artificial intelligence content writing possible?

We’ll explore all of these questions, but first, some definitions.

An oversimplified definition of Artificial Intelligence is that it’s the ability of machines to perceive, synthesise and infer information – just like our human brains do. In practice, there are many fields of AI, such as machine learning, neural networks, robotics, and natural language processing (NLP).

We were lucky enough to speak with Associate Professor Ali Mirjalili, Director of the Centre for Artificial Intelligence Research and Optimisation at Torrens University Australia for the Research That Matters podcast. Professor Mirjalili was recently named the world’s top AI researcher in The Australian’s Research 2023 magazine.

In a field of AI that focusses on natural intelligence, Professor Mirjalili takes his inspiration from nature.

“If you leave some food on your kitchen counter ants will eventually find it and it’s been proven that ants can find the shortest path between their nest and food sources. Imagine a truck is an ant; for the truck we want to find the best path, the optimal path, between warehouses and the customers,” explains Professor Mirjalili.

ChatGPT is based on Large Language Models (LLM) another sub-field of AI that observes our conversations and articles on the internet. ChatGPT stands for Chat Generative Pre-trained Transformer – what that means is the AI has been trained by observing conversations between humans.

The New York Times podcast Hard Fork explains (about 5 minutes in) that ChatGPT has analysed hundreds of billions of examples of text from across the internet. It breaks those words down to units of text and works out which bits of text belong with other bits of text in certain contexts.

So, when you ask ChatGPT a question it responds with the token of text that is likely to come next, based on the tokens that have come previously in the conversation.

In my explanation of how ChatGPT works, I’ve paraphrased a conversation from the Hard Fork podcast. I’ve also referenced Hard Fork and provided a link to its home on the New York Times website.

ChatGPT doesn’t do that.

In a thoughtfully written blog, musician Nick Cave describes ChatGPT’s output as ‘replication as travesty, or a kind of burlesque.’

“ChatGPT’s melancholy role is that it is destined to imitate and can never have an authentic human experience, no matter how devalued and inconsequential the human experience may in time become,” laments Mr Cave.

In addition to passing off other people’s work as its own, ChatGPT has a tendency to lie.

Christopher Graves is the President & Founder, Ogilvy Center for Behavioral Science at Ogilvy and he asked ChatGPT to cite some scientific studies. The chatbot cited several research studies which Mr Graves couldn’t find – because ChatGPT made them up.

Mr Graves asked ChatGPT specifically not to fabricate, but to deliver citations from studies that actually exist. It replied with three more authentic sounding journal articles, which don’t exist.

Even though it can’t cite sources accurately, schools and universities across Australia and around the world have rushed to ban and block ChatGPT from running on their networks.

Meanwhile, Google has declared content automatically generated with AI writing tools to be spam. Thanks Google!

On the other hand, Google is working on its own Large Language Model chatbot that will provide more source attribution – which could generate revenue via pay per click (PPC).

The ethical challenges of ChatGPT are not the trials of it alone, but the struggles of companies who interact with it through APIs and individuals who apply its immense power to solving problems large and small.

Many of us remember the arrival of the smartphone, the desktop computer and even the cordless phone which allowed us to take calls away from the prying ears in the living room. Like all of these tools, the imminent AI boom will bring an exponential increase in ethical challenges that we must consider.

Hands up, who thinks that you can find all the information that you want on the internet if you know how to use a search engine effectively? I hate to be the one to burst your bubble, but you’ll only find information that’s been indexed by a search engine and prioritised by its algorithm.

Think about it. When was the last time you received a search result in a foreign language? Google does a great job of translating, but have you ever found the best answer to your question in Chinese, Spanish or Arabic – the most common languages online after English.

Google tends to return results in the language you’re searching with, which for me means English results from ‘The West’. If generative chatbots are trained on the same English material, they’re likely to reflect the social norms, ethical values, traditional customs, belief systems, and political systems of the Western World.

Knowing that ChatGPT was exposed to training data with a cut-off of 2021 and may not know about less well-known people, I put it to the test to find some diverse voices. While it was aware of Antoinette Lattouf, Santilla Chingaipe, and Kon Karapanagiotidis, ChatGPT had no idea who Kemi Nekvapil, Winitha Bonney OAM and Rana Hussain are.

Future versions of ChatGPT might learn from even bigger Large Language Models that draw on more languages and diverse voices. But there’s another diversity issue.

If you and I both ask ChatGPT to write a well-researched blog on whether artificial intelligence content writing is possible, we’ll both end up with very similar content on our websites.

It’s what Ian Whitworth calls the Penguin effect. Our website is a penguin in this allegory and it’s a very attractive penguin, but when you zoom out to the horizon, it’s just one of an infinite waddle of identical penguins (websites) squawking.

At Written & Recorded, we research broadly for our copywriting and podcasts to include a diverse variety of views, voices, and verification. While AI might know that there are known unknowns and unknown unknowns, it still answers with authority based on the data it’s been exposed to – while we think outside the box and ask more humans more questions.

Now, let’s take a closer look at how artificial intelligence content writing is possible without decreasing diversity or increasing ethical challenges.

It’s important when using AI to write content that you choose your bot carefully and keep a close eye on it – for more than just ethical reasons.

We spoke to a client recently who was seeking support for her social media posts. Her main concern was that we humans would actually be writing the posts and monitoring the social media accounts.

As we dug deeper into her robot fears we found that she had engaged a content agency in the past who relied on AI or automation to post for her. It appears that the automation was looking for basic keywords around birthdays, anniversaries, or life events to provide a basic message along the lines of “I’m so happy for you”.

Unfortunately, it found the keyword it was looking for in a post about the death of a family member from one of her clients. Needless to say, “I’m so happy for you,” was not the preferred response.

The comment was deleted but because of the nature of social media, it could still be seen on the original post and it haunted that business for a long time.

With those cautions in mind, here’s three forms of AI that you can use to write content.

Bad jokes aside, ChatGPT is great fun to play with. We had a great time asking the bot to explain podcasting in the style of Donald Trump, Jerry Seinfeld, and Homer Simpson.

We also asked it to compose a song about podcasting in the style of Lou Reed, LL Cool J and N.W.A.

ChatGPT clearly has far more potential than just using it to write your content. We think that its use of probability to work out what word comes next in a given context is an opportunity to get inside the head of your target audience.

Why not ask ChatGPT to tell you three things that are on the mind of the people you want to engage with? You can use the results to create your next three LinkedIn posts, blogs, or even to build out an annual content calendar.

We asked ChatGPT to identify three things that are keeping Chief Marketing Officers (CMOs) awake at night and got three great topics to start working with.

There are two great advantages of using automated transcription to write content.

Firstly, you can get your ideas down fast. Simply use your phone to record what you want to write about, then upload the file to an automated transcription tool like Sonix. You might be surprised how quickly you can get some pretty workable words on the page.

The second major benefit of this method is that you’ll create content that sounds like you’re speaking, rather than content that sounds like you’re writing. It’s a great hack for creating engaging content that leaps off the screen and captures your audience’s attention.

Once again, it’s important to keep a watchful eye on your AI and proofread carefully. We often have a laugh at the AI transcription service that we use for our interviews.

It does a reasonable job if we speak clearly and provide a very high-quality recording. But most people have a bit of an accent, or an unusual way of speaking – and being on the phone doesn’t help either.

“That’s alright but I think that’s great. Yeah, I don’t have a lot of space to write home actually writing to 500 words so they’re pretty tight. The Ladoga school is heaps but that’s where I’m at.”

That’s me explaining to a client that I’ll be writing a 500-word blog. It never stops making us laugh – Thanks AI.

That said (or transcribed), it’s been really helpful for remote collaborative editing on the Competition Lore podcast.

Full disclosure, we don’t use Grammarly, but we know people who love it!

Unlike the Large Language Model of ChatGPT, Grammarly uses advanced machine learning and natural language processing to improve your writing. Of course, it will pick up typos and suggest grammar corrections, but it also calls out tone-of-voice issues – so you can adjust your writing to be less aggressive, or more confident.

The free version of Grammarly will help you spot typos, write more succinctly, and give you a rating on your tone. The premium version will actually rewrite your text with more spectacular vocabulary, clearer sentences, and a confident tone.

It sounds like it could be a great help when preparing text, but it probably won’t spot incorrect names of people or businesses – so be sure to proofread your AI proof-reader carefully.

The key element in content marketing is connection with your customers. You need copywriting written by humans, for humans that answers the questions that your customers are asking.

There are some really helpful AI tools for sharing content, but you have to be ready to respond like a human. Your customers are likely to connect with your brand several times before converting. They’ll be more likely to convert if those early connections are warm and friendly, rather than cold and calculated.

In their article 12 Social Media Mistakes That Will Drive Followers Away From Your Brand, Forbes highlights failing to listen to customers, social presence with no personality, and not engaging with your audience as big no-nos.

AI can help you avoid these mistakes, but it can’t do all the work for you.

How can you tell if you’re reading content created by a human? Well, if you have to ask…

Writing this blog has been a rewarding and joyful experience for me – feelings that I have tried to convey to you, the reader, with my words, structure, and links to so many interesting sources.

At Written & Recorded we bring the same enthusiasm, curiosity and wonder to the copywriting we do for our clients. We are humans, writing for humans.

To be human is to feel a connection with other humans. When two apparently incongruous ideas are presented together in text and the reader makes the same connection as the writer – that is the power of great content.

Artificial intelligence content writing is possible, but humans can make it more engaging, effective, and exciting. Let us show you how.

A copywriting service like Written & Recorded can write unique content that is highly researched and extremely engaging. So, you can be sure that your audience will read and benefit from your content – whether they are humans, or search engines.

We’re a premium copywriting service that can convey complex ideas in words that are understandable and enjoyable. Drop us a line in the contact form below or call 03 9523 6196 for a chat about creating the content that your customers are searching for.