During the pandemic, seeing an AI-generated image on social media would be immediately flagged as fake, but there were only a few pieces of content created with artificial intelligence with the purpose of confusing people. However, following the end of the pandemic and the boom of AI, the technology evolved so much that basically anyone could create a difficult-to-identify fake image.
UNESCO states that AI literacy is now necessary to be able to distinguish the real from the fake, so media literacy alone isn’t doing the job. And while this applies to visual content, written texts are also becoming, in the majority, more artificial due to the use of these tools. So, while you’re reading a post on social media, you may be unable to recognize that it’s generated by AI, since the tone and language copy almost perfectly the human one. But how can you identify genuine messages and texts without losing it? Let’s find out.
Texts generated by AI are increasingly of high quality
If the beginning of AI writing could’ve been described as robotic, forced, and full of cliché phrases, the tools with which you can generate texts are continuously evolving, so they’re adapting to how a person would actually write or talk. Therefore, if someone trains the AI well enough, these assistants can craft genuine texts.
However, what’s interesting about these technologies is that when you feed them too much data, they can start making up things. There’s something called “chat fatigue” observed in instruments like ChatGPT or Claude that happens when the technology is (literally) tired, so it starts hallucinating and giving you made-up facts, especially if the language used was other than English.
So, how do you analyze AI language?
If you’re a student or an employee having to analyze huge amounts of articles or information (such as a journalist or social media manager), you will encounter AI content more often than not. Being able to seize it can save your academic life or your career, because no matter how important AI becomes for humanity, reading an AI slop text isn’t doing any good for clients.
Considering that ChatGPT is the most used AI tool at the moment and is accessible for anyone without prior knowledge about how to optimize it, the safest solution is to use a ChatGPT detector that is trained on multiple LLM models. This tool analyzes your text and offers you a thorough report on the level of plagiarism in the text while highlighting the phrases that get close to AI writing and the percentage found in every sentence.
Luckily, you can use AI to detect AI, and software like the one above are trained on other models like Claude, but the latter is known to be less efficient. Some people consider that ChatGPT is ideal for execution while Claude is made more for communication, but this discussion gets deeper the more we dive into it.
On training yourself
Training yourself by reading AI texts in social media posts or news is an essential step of your AI literacy education. Sometimes, identifying such writing is considerably easier when the person behind it is a novice, because they only copy-paste what the AI generates. Such texts include things like an emoji at the beginning of every sentence, the overuse of the dash, and the continuous repetition of the same idea without getting anywhere.
The next level of recognition is texts that contain predictable word choices, generic explanations (instead of personal experiences), and sometimes a really weird choice of words when simple terms could be used. If you don’t know where to start with exercising, know that a huge amount of AI books are making their way into libraries, so you can read one and see what we’re talking about.
Circle back to fact-checking
Fact-checking is always recommended by academics and specialists, but the process might be too lengthy for internet users whose attention span is no more than 40 seconds. However, this simple activity can literally save the world, because it makes you, as a reader, question the text and search for the information presented to ensure it’s real.
Fact-checking is the strongest tool against misinformation. When someone uses AI to write a text without caring much about it, it might happen for the AI to cite sources that are either unreal or are irrelevant to the message. This is why checking the links that AI uses as evidence can tell you straight away if the text is original or not.
Twisting terms and truths is a common practice during political events, so this is when you should be most careful about what you’re reading. If something seems sketchy, research it immediately to ensure you’re not fooled. But this also requires a thorough understanding of the world.
Learning AI literacy is already a thing
Who could prepare future generations for the rise of AI more than universities? Stanford University created an AI Teaching module that includes levels of AI literacy: functional, ethical, rhetorical, and pedagogical. The framework is pretty dense and includes valuable information for anyone to check out.
You will find official sources like this diving deep into the AI literacy subject, and you should take advantage of them, because this is how you will navigate the future of the internet. You will be able to make informed choices as a regular user, but as an employee, this skill can support more productivity at work by combining AI with human intelligence more efficiently. AI literacy will be needed across industries, especially because AI-based tools will become essential in improving their workflow.
What do you think about AI literacy?
As AI develops and is capable of generating high-quality images, videos, and complex messages, we need to be equipped with the tools and knowledge of distinguishing such media from human-based ones. When it comes to articles, news texts, and social media advertisements, recognizing AI texts is necessary. You can use a mix of AI detector tools and put your knowledge to the test when reading these texts and make the right choice to believe them or not.
The post Why Is AI Content Harder to Identify and How to Fact-Check Everything Like a Pro appeared first on Daily Emerald.