Exercise caution when using AI-generated content

ChatGPT offers tenuous suggestions, not true solutions

9/14/23

A popular program that makes artificial intelligence (AI) available to the public has polarized content creators.

ChatGPT, the AI tool leading the charge for generative machine learning, looms at the center of this binary construction in which some marketers eagerly embrace the premise of AI-driven copywriting and others caution against a promise too good to be true.

Whereas chatbots have been trying to mimic human conversation for years, ChatGPT presents an evolved, versatile platform that anyone can use to write business pitches, compose fiction, churn out essays and articles, and summarize and translate text. Obvious applications for marketers include blog posts, website/SEO copywriting, whitepapers, social media posts and ads.

The benefits of adopting ChatGPT and programs like it are readily apparent: the speed to produce text, scalability, opportunities for personalization, consistency and cost.

AI-generated content extends beyond the assembling of words to include images, audio, video and even code. That’s right: programs can write programs, including code with robotic applications, extending the algorithm’s reach from the virtual space to the real world.

And tech experts believe this is only the beginning.

Context

ChatGPT isn’t the first iterative AI author. Ever since the early 1980s, personal computers have tried to help human writers be a little better by offering spellcheckers — built-in tools that have since evolved to also cover grammar errors and other opportunities to make one’s writing more concise, free of clichés and generally less awkward.

In addition to editing, ’80s AI didn’t shy away from the heavy lifting of writing. In fact, AI explored new territory in 1984 with the composition — and publication — of a book titled The Policeman’s Beard is Half Constructed.

Is that collection of poems, short stories and other miscellaneous scenes a literary masterpiece? Heavens no. It reads like HAL from 2001: A Space Odyssey narrating a virus-induced fever dream. Nevertheless, it proved that algorithms could approximate original human writing.

During the past decade, rapid advancements in machine learning and natural language processing have led to considerable increases in the capabilities of these content-creation systems. Simply put, AI has taken greater strides to impact the words we write and read. In 2007, StatSheet created content related to sports statistics for online audiences. Two years later, Grammarly entered the scene, and the program continues to make a big impact on writing that ranges from academic assignments to mundane emails.

Earlier this year, ChatGPT’s public debut made an immediate impact on the marketing scene, suggesting AI is here to stay.

Controversies

When technology surges forward, concerns inevitably bubble up in the wake. Here’s a summary of some of the biggest problems AI-generated content presents:

Google prioritizes people

Nowhere has Google stated that its own algorithm penalizes AI-generated content, but recent updates to the search engine somewhat imply it. Google’s “Helpful Content” approach firmly puts people first, granting favor to articles and other content written by established subject matter experts.

ChatGPT’s lack of a byline puts it at a distinct disadvantage. Furthermore, longwinded compositions that feature more SEO keywords than useful details have gone the way of the answering machine. In other words, just because ChatGPT can manufacture paragraph after paragraph of fluffy prose, that doesn’t mean you should use it.

Savvy marketers are holding off on populating their websites with machine-made copywriting — at least until Bard, Google’s AI initiative, joins the party.

ChatGPT lies

The technical term is “hallucinate,” but the result is the same: ChatGPT evaluates everything that is fed into it, including opinions and outright falsehoods. The internet — which makes up the bulk of ChatGPT’s education — presents a mixed bag of information. Even though the program attempts to pluck facts from the pool, its judgment remains lacking, resulting in scenarios where ChatGPT gives two different, contradictory answers to the same question.

Worse, ChatGPT doesn’t reveal its sources, which makes fact-checking cumbersome at best.

AI-generated content is essentially stolen

ChatGPT cannot create; it can only collect and customize. Yet that content comes from somewhere. Succinctly put, ChatGPT snatches whatever it needs from wherever it pleases without permission or credit.

Millions of content creators’ work is mined daily to fuel ChatGPT’s tireless constructions. Worse, AI has gotten so good at mimicking established voices that books have been appearing on Amazon.com with the parroted author’s name on the spine when, in actuality, the book was composed by a bot — with an all-too-human scammer pulling the proverbial strings.

AI makes cheating easy

On the other side of the equation, teachers and professors are scrambling to find ways to thwart plagiarism since it’s now all too easy for ChatGPT to write papers and complete other academic assignments for students in seconds.

The free program can also answer most test questions, but results vary greatly. In the same way that ChatGPT lies, it also frequently fails the test.

AI-generated content perpetuates other existing problems

Just as ChatGPT filters the truth along with the lies, the program repeats mankind’s bad habits, including instances of bias and prejudice. Programmers are already hard at work to prune the learned behavior that leads to mimicking humanity’s worst qualities.

Accidents are one thing; intentional exploitation, quite another. ChatGPT is more than capable of writing and disseminating phishing emails and malware when combined with OpenAI Codex. The influx of spam in your inbox and the proliferation of fake profiles posting on social media will only get worse. To quote one cybersecurity professional, “The barrier to entry is getting lower and lower…to be hacked and to be phished.”

Businesses have landed on opposite sides of the AI line, with some outlawing the use of AI-generated text by staff and freelancers. Meanwhile, AI detectors are evolving as quickly as the AI-generation tools themselves.

Over the past few months, more than 20,000 people, including leading computer scientists and tech founders, have signed an open letter calling for legislation to stop — or at least delay — giant AI experiments like ChatGPT. One politician has warned that AI could result in not only cheating, job losses, discrimination, and disinformation, but also uncontrollable military applications.

Some of the above concerns sound like the trappings of science fiction, where self-aware machines assert their will upon us mere mortals. But ChatGPT and other bots have no self-made agenda. Whatever harm comes of them is far more likely to be the result of pursuing misaligned goals that stem from human error or the direct consequence of people using the tool with intentional malice.

At the moment, however, ChatGPT is merely a free, feasible tool for individuals and businesses to harness however they see fit.

Conclusion

ChatGPT isn’t going away, and before long it will be joined by rivals such as Google Bard and whatever “X-citing” alternative Elon Musk cooks up. Microsoft has already announced plans to integrate OpenAI into its search engine, internet browser and other software products.

Growing pains and learning curves aside, there are appropriate ways to implement emerging AI tools into your content-creation strategies. In its current incarnation, ChatGPT can serve as an assistant but shouldn’t be given the keys to the corporate kingdom.

Early evidence suggests it’s better to use ChatGPT and its contemporaries for early planning and other exercises that don’t face the public unfiltered, such as:

To help explain concepts at different difficulty levels
To provide feedback on your original writing
To get suggestions for marketing topics
To generate metadata
To banish writer’s block
To outline pages and posts

All those assets, then, can be used to buoy human-written, heartfelt, authentic thought pieces that speak directly to the intended audiences — efforts that admittedly still require time and money.

Because if you trust a free tool implicitly, copying and pasting without scrutiny, you’re bound to get exactly what you pay for — or, worse, make a costly mistake.