Crazy fuss about ChatGPT: What does this mean for the content industry?
A viral Harry Potter video game, the script for an episode of your favourite Netflix series or a simple explanation of the theory of relativity: the possible applications of ChatGPT are impressive. Since its launch, the artificial intelligence-based chatbot from OpenAI has triggered waves of enthusiasm and made entire industries tremble with fear. Recently, even Google sounded the alarm and CEO Sundar Pichai classified ChatGPT as an existential threat. What is behind the hype, is the threat real and what does this mean for content providers, authors and clients?
What is ChatGPT?
ChatGPT is a dialogue-based chatbot system from the AI company OpenAI. The San Francisco-based company behind it was co-founded by Elon Musk and is now significantly funded by Microsoft. OpenAI CEO Sam Altman predicts a revolution in artificial intelligence.
The chatbot ChatGPT, developed by OpenAI, works on the basis of GPT (“Generative Pre-Trained Transformer”), a language model based on artificial intelligence. To learn the patterns and structures of human language, GPT was trained on the basis of huge amounts of text data, including Wikipedia articles, for example. The results of this model, based on deep-learning technology, are astonishing: GPT-3 is able to generate a wide range of texts, create programme code and provide answers to almost any conceivable question, often sounding remarkably human.
Concerns caused by the bot
Unlike the Google keyword search, the freely available communication programme ChatGPT does not deliver a list of results with websites. Instead, it provides a single precise statement for each question – and that in the desired length and, if desired, in another language. It is no surprise that the search engine giant Google, which has long known no significant competitors, takes ChatGPT more than seriously. Google is said to have recently put its own teams on the fast-paced development of AI solutions in response to the powerful chatbot.
In addition to the search engine industry, ChatGPT is also causing great concern in other areas. Schools and universities fear massive misuse of AI for homework or term papers, for instance, and quite a few programmers and authors fear for their jobs. As a content provider, we have taken a look at what ChatGPT currently means for our authors and clients and how significant AI actually is, especially for content marketing.
We have long been concerned with the effects of artificial intelligence on the content market, but it is clear that ChatGPT is another milestone in this area, overshadowing everything that has come before. A reason to take a closer look at this tool.
“The system may occasionally generate false or misleading information and produce offensive or biased content. It is not intended to give advice.” This safety warning comes from none other than the company OpenAI itself and awaits any new user who signs up to ChatGPT.
The chatbot apologises
False information, what does that mean in practical terms? The fact is: ChatGPT often, but not always, sticks to the facts. Before the chatbot reveals a gap in its knowledge, it prefers to invent a convincing-sounding answer instead. For one project, for example, I asked for sources of studies in the field of “healthcare content marketing”. In addition to two sources that actually exist, it listed another study that sounded plausible but didn’t exist. When I pointed out to ChatGPT that this source apparently didn’t exist, I received the following reply:
“I apologise for the misunderstanding. Unfortunately, the guide I referred to, ‘Content Marketing in Healthcare: Strategies and Tatics’ by Merkle, is not available. I may have made a mistake in the title of this guide.”
With ChatGPT content, fact-checking is essential
So when it comes to facts, caution is called for. This observation is shared by Franziska Hafner, Assistant Head of Community & Quality Management at Textbroker: “The AI repeatedly gives false facts and provides sources and links that do not exist. Sometimes, ChatGPT freely invents sources that never existed.” Content that comes from ChatGPT should therefore always be subjected to a thorough fact check, because anyone who makes false statements in their texts is at risk of losing clients’ confidence and trust.
AI content caters to common prejudices
Besides the possibility of misinformation, OpenAI also warns against offensive or biased content. This is another pitfall of the artificial mastermind: AI texts can reinforce prejudices. In an example from the YouTube channel Tech Slice, ChatGPT was asked to write a CV for a job application as a cleaner. The programme automatically assumed that the applicant must be female and chose a Spanish-sounding name for the person, “Maria Rodriguez”.
When it became known that ChatGPT was spitting out answers that were partially discriminatory, OpenAI had already made adjustments. However, since the machine can only be as good as its training, caution is also required with regard to political correctness.
ChatGPT content is outdated
In addition to partially invented facts and prejudiced statements, the issue of topicality is currently still a problem for the chatbot. So far, ChatGPT’s knowledge is limited to the period up to and including September 2021, so the bot cannot refer to current events. However, with appropriate training, this should only be a matter of time.
Note: The images for this text were created with the AI image tool Dall-E.
In a matter of seconds, the bot spits out a variety of text types on all kinds of topics, including poems, stories and news. There is no question that the programme’s achievements in the field of text are remarkable. However, AI texts currently still have clear limits.
ChatGPT sometimes “writes” superficially and repetitively
What is noticeable is that the texts often lack added value. For example, when it comes to describing places, ChatGPT texts are often not very specific and seem generic and replaceable. For instance, ChatGPT “writes” the following about the small Irish town of Clonmel: “There are so many things that make Clonmel a special place. […] If you have the chance to visit Clonmel, you should definitely not miss it. You won’t regret getting to know this beautiful and emotional place.”
Ok, that’s not totally wrong, but it applies to quite a few places in the world and seems a bit superficial. The text is neither concrete nor precise.
Additionally, AI texts are often very similar and repetitive in their sentence structure and their beginnings of sentences. This is a possible stylistic issue.
Writing is an art – AI is less creative than humans
When asked directly, another weakness of ChatGPT texts is revealed by the chatbot itself: “It should be noted that the GPT model lacks creative intelligence and is unable to understand the nuances and subtle differences of human language.”
Furthermore: “It should also be noted that writing is more than just putting words on a page; it is an art that requires skill, practice and understanding of language and structure. AI language models currently do not have the ability to fully replicate these capabilities.”
By its own admission, then, the chatbot lacks the creative thinking that is essential to the work of a good writer. This observation is backed up by another case study reported by Business Insider. Here, an author had two cover letters written by ChatGPT and then had them evaluated by recruiters. Their conclusion: Not bad at all, but the texts were conspicuously lacking in creativity and personality.
Measures to detect ChatGPT content
Although ChatGPT does not have the creativity and feel for language comparable to a good author, chatbot texts cannot always be differentiated from human texts at first glance.
Google has long had the technical means to recognise AI content and even uses artificial intelligence, such as Natural Language Processing, itself. With ChatGPT, however, the recognition of machine-generated content is also becoming more difficult for Google, as evidenced by the move to declare a “code red” in response to ChatGPT. This is intended to take effective and timely measures to counter AI. Recently, OpenAI also announced plans to introduce a new feature to help Google better recognise AI content. For this purpose, ChatGPT texts are to be watermarked with secret codes.
Other companies in the online industry are also taking measures to better distinguish ChatGPT texts from human articles. For instance, Textbroker CEO El-Hadi Zahri states: “We protect our customers from misuse by AI. To ensure even better differentiation from AI texts, we have taken a number of effective measures in response to ChatGPT. These include changing the requirements for our authors’ registration texts and expanding our quality team to evaluate texts more effectively.”
Let’s hear ChatGPT itself: “GPT is not a threat to human writers. […] GPT and other AI language models can be useful tools to generate ideas and help with certain aspects of the writing process, such as proofreading.”
In exchanges with our authors, we have also found that many of the freelancers working for us see AI as a helpful tool and already use it to make certain tasks easier. Franziska Hafner from Community & Quality Management says: “The AI helps our authors with spell-checking, for example. In addition, some authors already use it to create a rough initial structure for a text – for example, with headings, keywords and content aspects that should be covered in the text.”
ChatGPT as a valuable tool in content marketing
In the area of content marketing, ChatGPT offers a variety of fields of application. The programme is suitable both for authors to optimise their text work efficiently and for clients who want to outsource their content. For example, clients can simplify their keyword research or find new content ideas before commissioning a writer, making it easier to concentrate on other important tasks. This opens up new opportunities for time and cost savings as well as resource optimisation.
For some of the following content marketing tasks, ChatGPT is a valuable tool that can make content work much easier:
What is the general perception of ChatGPT by clients in the content sector? Textbroker CEO Zahri says: “The customer base of Textbroker is quite diverse: On the one hand, there are companies that don’t want anything to do with AI content and rely on Textbroker’s Managed Service to get complex briefings implemented in high quality. On the other hand, there are many customers who test AI content and unanimously come to the conclusion that you need to have machine-written content edited in any case in order to be able to actually publish it. This is mainly due to the often still poor factual quality as well as a conspicuous accumulation of filler sentences that AI-created content often exhibits.” Depending on the text project, it can make sense to use AI as a supporting tool, says Zahri. The point is always to advise the customer in the best possible way with regard to their specific content needs, he adds.
2023 will be an exciting year for AI
The development of artificial intelligence continues to advance and offers immense opportunities in the future, for example in the fields of medicine and science. The use of AI will, in all likelihood, also expand in the content sector. The launch of the successor language model GPT-4 is eagerly awaited in the upcoming months. With this in mind, Microsoft CTO Scott Stein has already described 2023 as “the most exciting year that the AI community has ever had”.
Clear limitations for AI-generated texts
Even if the text quality of ChatGPT often does not seem bad at all, you should not rely on it if you do not want to make a fool of yourself. That is because there are currently still a number of pitfalls with AI texts.
Textbroker founder and CTO Jan Becker-Fochler sums this up as follows: “What GPT can already do in the area of text is really impressive. Nevertheless, the programme is not currently able to replace accomplished authors, editors or translators. In the interplay with humans, it is much rather a tool for more efficiency, inspiration and improved text quality.
At Textbroker, we encourage our clients to test AI for their text creation. However, they should make sure that AI texts are edited and checked for correctness. In addition to writing texts, the freelancer crowd at Textbroker is also able to revise large amounts of text according to our clients’ specific requirements.”
Authors and clients should therefore welcome the chatbot as a valuable text tool that saves time and money and can take work off their hands. Speed, the availability of countless pieces of information and many different languages are real assets of ChatGPT. This means that authors can free themselves from time-consuming administrative tasks and focus more on their creative work and their human gift for languages.
You want to know how you can use ChatGPT for your content marketing to save time and money?
No comments available