ChatGPT in Technical Writing: Risks and dangers
The usage of artificial intelligence has revolutionized the approach to performing dozens of various jobs. You no longer need to feel your head dizzying when composing an email to your business partners, suffer from a “fear of a blank page,” or go down in a world of controversial information in a particular field. Research, analysis, brainstorming ideas, and some basic content creation can now be easily performed by modern AI tools.
However, is everything working as well as expected? Can artificial intelligence substitute most of the Creative and Technical Writing in your daily working routine? The hidden truth is that AI still can’t perform many tasks with a quality similar to humans. Here is a list of risks you need to know when putting AI-generated solutions into practice.
Outdated data
Like human writers, artificial intelligence needs to do research when completing a particular task. It will automatically gather different types of data from numerous sources, analyze them, and create a relevant reply to your request. However, in this case, choosing only 100% trustworthy and up-to-date sources is the biggest issue.
Here comes the first weakness of AI-generated content — you never know what sources it used to complete your response. Some of them might turn out to be irrelevant, while others are not reliable. In some cases, AI tools miss important numbers and statistical data.
Here is an example of inaccurate content generated by ChatGPT. I asked it to provide the address of the SoftServe office in Wroclaw. AI’s reply was the following:
But if we check the most up-to-date information on the company’s official website, we will discover that the Wroclaw office has another location:
This mistake might not look like a critical error. But imagine sharing more important but irrelevant information with your clients or partners. You face significant risks of losing your expertise in the eyes of the readers, which can cause reputation damage to your company.
This issue is one of the most important reasons why most companies prefer hiring Content and Technical Writers. Real writers can’t be substituted by even the most innovative tool or utility, including AI-based solutions.
Inaccuracies
You can hardly find professional writers who don’t know what to write about when having a specific topic. As for AI-generated content, you might suddenly face an issue when you receive an output that is not fitting your main idea. Even if you provide an AI tool with a plan that contains a list of points to describe, the result you receive might still not be satisfying.
Here is an example of when ChatGPT fails to meet your instructions:
First, AI didn’t generate a 3-paragraph writing. Second, it uses the keyword five times instead of four.
Another issue is that ChatGPT often makes logical mistakes. For example, it might not notice the irrational system behavior described in the text, which is actually a bug.
And what if we add even more rules? The result will likely turn out to be even worse. Therefore, professional Technical Writing is still not a good fit for artificial intelligence. If AI fails to meet simple requirements, it will inevitably break down the whole structure of the most complex technical documentation.
Plagiarism
As a rule, AI generates fresh and relevant content that is nearly 98% original. However, it still has a high risk of composing plagiarized articles since all the information is collected from other online services. Unlike a real writer, AI can’t come up with clear conclusions, use real-life experiences, or share new facts or data. It is only good for processing someone’s else ideas.
So, what is the risk of sharing plagiarized content? First, there are legal issues: intellectual property rights and copyright violations. Second, is a drop in your SEO scores (Google and other search engines rate plagiarized texts as invaluable content that brings little to no use to the audience.)
Failure to meet the needed voice and tone
Writers can create content for completely different audiences, use (or avoid) complicated technical concepts and slang, as well as describe information from different angles. This is not about AI. In most cases, ChatGPT and similar tools create primitive content that fails to meet your voice and tone requirements.
Let’s imagine you use AI to compose pieces of technical documentation. Each time, AI will generate information according to the topic you choose. It will use various style guides, instructions, and approaches to writing. As a result, your document will look inconsistent and unprofessional.
Of course, you can make ChatGPT learn by sharing various style guides and examples with it. However, the final output of the AI-generated document still requires proofreading and editing by a Technical Writer.
Lack of creativity
The brain is still more powerful than artificial intelligence solutions. Of course, you will likely lose if you play chess with AI. You will not write a simple email faster than ChatGPT. However, when it comes to crafting quality content, nothing can be more effective than a human writer.
The reason is that AI has a limited ability to create content that requires sharing personal experiences, out-of-the-box solutions, and uncommon insights. It is only good to follow basic requirements (still, don’t forget to proofread it — the mistakes might still occur!)
Here is an example of creative banners for ABC company proposed by ChatGPT:
The list of proposed options doesn’t look unique or eye-catching. It is possible to provide more instructions to AI and explore the possible alternatives it will generate for you. However, teaching it to be more creative will take you time (composing your own banner will likely be faster and more effective)
Security concerns
The use of AI technology carries the risk of personal data leaks due to those issues:
- Data breaches
- Algorithmic bias
- Inferential disclosure
- Membership inference attacks
- Model inversion attacks
- Third-party data sharing
- Inadequate user awareness
To mitigate these risks, organizations should implement strong security measures. Hiring an experienced Technical Writer who realizes the possible security concerns and can use AI-based tools wisely is a safe and reliable way to compose technical documentation.
The “robotic feeling”
Mostly, AI-generated content has a simple structure and writing style. Even if you add direct instructions to make it readable, interesting, and inspiring, the generated solution will still fail to meet your requirement.
The main issue here is that AI composes fully functional but forgettable content that will not encourage the reader to subscribe, make a purchase on your website, or just think for a couple of moments about the topic.
Although ChatGPT can be trained to create content with different voices and tones, it still desperately needs to be supervised by a Technical Writer. The reason is that it still often fails to meet your instructions and generates content that doesn’t perfectly fit your unique requirements.
As for technical content, such as user guides, installation guides, manuals, product specifications, and release notes, using AI might simplify all instructions and make them too short. As a result, the users will find them tricky and difficult to follow.
Over-dependence
Dependence on AI can result in declining critical thinking, problem-solving skills, and attention to detail. ChatGPT does not highlight the errors and show a correct option on the side panel (like Grammarly does), so you can’t learn from it and compare before/after the AI review.
Conclusion
All in all, providing reliable, understandable, and up-to-date information is crucial for both Technical and Creative Writing. AI might help writers to automate some basic workflows, but it still can’t substitute them.
First, deep analysis, setting up clear document structures, and using step-by-step writing schemas can be performed only by real experts, not tools.
Second, appealing to emotion, grabbing attention, and engaging the reader to perform a certain action is a writing advantage of human writers only.
To get the most out of ChatGPT, you should get used to working with it, learn how to build prompts, be precise in setting tasks, and be careful not to expose sensitive data. And, of course, remember to always double-check the generated information. Meanwhile, AI tools are not the only risky thing in a Tech Writer’s job. If you want to find out more about mitigating risks, check my colleagues’ article on How to handle risks in Technical Writing.