My big pet peeve is that it was a major mistake calling LLMs artificial intelligence in the first place, because it engenders it with anthropomorphic qualities (e.g. a will, feelings etc.). It would be more accurate to call the present models for smart algorithms. At present there isn’t any model that is close to approaching intelligence. Another thing that often appears difficult for beginners is that how and in what order you present a word can heavily change what type of outcome you will receive from your prompt because of the token weights and the associative power different words play in a LLM-model. And even if you master the right prompt, you won’t necessarily achieve the same results from time to time because of the built in elements of randomness that LLMs have. I think a basic understanding of an LLM would be a plus. Just a schematic one will bring you far. Thirdly – well, for every task you assign, you need to assess what a wrong answer will involve, and to what extent you would be better off doing the task yourself, if it means you need to fact check the result to mitigate the risk. This again engenders a necessary base knowledge in relations to the task – i.e. if you ask an LLM about quantum physics be certain that you know or can find alternative resources to be able to assess the LLM-results before submitting your ‘findings’ for a seat at NYC Quantum Summit
I guess the main problem would be the fact of not understanding what generative AI is. As the interface looks like a regular chatbot and the entry difficulty is near zero, there is a risk of not recognizing the pattern in how AI works and becoming something close as a total-trust-source. This belief would erase the border between the programme itself and its use, resulting in many of the problems we have been detecting in the last year: AI induced psychosis, parasocial relationship with it or plagiarism, false information or even hoaxes. Any new user must know clearly how the AI ecosystem works, keeping in mind this tool is not a miracle and is not a pocket with access to all internet knowledge. Without understanding what is a prompt, how the embedded words system works or how the model has been trained, it will be difficult to provide the information any user should give in order to get any desirable outcome and also, what kind of answer the AI system can offer. We could add the ethical and ecological issues AI has embedded to its use. They are not as important as the ones previously mentioned due to the fact they don’t affect the results of his personal use but I consider everyone should hear about them if anyone is resolved to operate this type of tool.
One of the most common mistakes people make is taking everything that artificial intelligence says as if it were an absolute truth, without questioning it. Often, the information provided is not cross-checked with other sources, which can lead to misunderstandings or the assumption of incorrect data. This can be particularly risky in situations that require accuracy, careful analysis, or informed decision-making. For this reason, the advice I would give to someone who is just beginning to use AI is to take the information “with a grain of salt.” In other words, it should be used as an initial reference point, but it is always essential to verify it against other reliable sources. Maintaining a critical attitude is crucial, and one should avoid becoming completely dependent on a single tool. By doing so, users can benefit from AI as a helpful support while still developing their own reasoning, judgment, and ability to evaluate information independently.
Interestingly, two things can be observed time and again among beginners. They simply use AI incorrectly and are then disappointed with the results. Most disappointments in dealing with AI are not caused by the technology itself, but by false expectations or usage patterns. Users almost always provide vague prompts and are then disappointed with the superficial results. The following applies here in particular: The quality of the answer depends on the quality of the question! AI tools such as Chat-GPT, Microsoft Copilot and Google Gemini are not designed as ‘search engines’ but as dialogue systems. It is therefore essential to think in terms of conversations rather than search queries. You should ask questions, refine answers by asking further questions, request examples and change perspectives. Only then will the results gradually improve. Artificial intelligence is an assistant that can sometimes make technical errors or even provide incorrect facts. It is therefore important to read the results several times, check the facts and personalise the content. AI tools work with probabilities, not with real understanding, so these systems provide drafts and food for thought, but not absolute truth. Even if the results are very impressive, these tools are not infallible. Before using an AI tool, you should ask yourself: Do I know exactly what I want? Have I clearly formulated my goal? Do I at least have a rough idea of the result in advance? Note: AI enhances clarity – it does not replace it! Is my prompt specific enough? Does my request include the target audience, context, style, length and desired output format? As mentioned above, the more precise the information, the better the result. Once I have the first result, I ask for alternatives and question any ambiguities. Good AI use is a dialogue, not a one-time click. Finally, check the result again, of course. Are the facts correct? Does the writing style suit me?
In my opinion, a common mistake made by beginners in dealing with AI is that instructions are too vague. If you only ask in general, you will usually get superficial answers. It is better to formulate clearly: Goal, format, target group and desired style should be described as concretely as possible. Another mistake is the expectation that AI will always deliver correct or perfect results. AI can also be wrong or invent content. It is therefore important to critically examine results and to improve or ask questions if necessary. Many beginners also underestimate the importance of iteration. Instead of simply accepting the first answer, it is worthwhile to sharpen it in a targeted manner: "Make it shorter", "Explain it more simply" or "Give an example". This significantly improves the quality. A typical problem is also holding on to a chat for too long. Once the AI goes in the wrong direction, it can be hard to correct it. In such cases, it often helps to start a new chat and formulate the request more clearly. After all, many forget that AI is a tool and not a substitute for one's own thinking. Those who actively cooperate, remain critical and steer in a targeted manner achieve the best results.
When using AI, beginners often make the mistake of giving instructions that are too vague. If you only write "Make me a text", the result usually remains general or inappropriate. It is better to clearly state the goal, style, length and target group. A second common mistake is to trust the first answer immediately. AI can be helpful, but it also makes mistakes or sometimes invents information. That's why you should always check important content, especially with facts, figures or sources. Another typical mistake is to give up too quickly in frustration when the first result is not perfect. Good results often only come about by asking questions, reformulating and gradually improving. So AI works best as a dialogue, not as a one-time command. Some beginners also expect the AI to "think" even though it only reacts to what is entered. The clearer the instruction, the better the result. These mistakes can be avoided by working with simple, concrete prompts, reading results critically and using AI in several steps. It is also helpful to give the AI a role, for example: "Explain it as if it were for a student" or "Write objectively and briefly". So you quickly learn how to get better answers. The most important point is: try it out, improve it and see AI as a support – not as an infallible solution.
One of the most common mistakes is applying the old “Google mindset” to AI. Many people are accustomed to entering just a few keywords into a search engine and then gathering the relevant information themselves. With AI, however, this approach often leads to disappointing results because the requests are too vague, too brief, or lack sufficient context. If you provide only two keywords, you should not expect a highly tailored or nuanced response. Another challenge is that many beginners are not yet familiar with the full capabilities of modern AI systems. If someone has never seen a more sophisticated prompt or does not understand the role of a system prompt, it can be difficult to appreciate what these tools are actually capable of. Working effectively with AI increasingly requires skills that resemble good leadership: clearly delegating tasks, defining objectives, communicating expectations, and critically evaluating outcomes. In most cases, the more precisely we communicate, the better the results will be. It is also important to develop a basic understanding of the differences between AI models and their approaches—for example, the distinction between fast, instant-response systems and more advanced reasoning models that take additional time to analyze and solve problems. Equally problematic is placing blind trust in AI-generated answers. Especially when dealing with important information, outputs should always be critically reviewed and independently verified. Another common mistake is dismissing AI as a short-lived trend. AI is not going away. On the contrary, the pace at which models are improving exceeds what many people imagine. That is precisely why it is important to take their potential seriously while maintaining a critical and balanced perspective.
I don't understand what beginners are asking about. They use AI for obvious nonsense, waste valuable time and resources, ask stupid questions, create really dumb images and videos without knowing how the creative industry works, and generally waste expensive resources on nonsense while polluting the earth—both figuratively and literally. That’s why maybe regulations should be introduced so that not everyone has access to AI, just as not everyone can be a nuclear engineer, doctor, astronaut, train engineer, etc.
The most common mistake beginners make is that they don't introduce the AI to the idea they have. Instead, they directly tell it what they want without any context or background. This is like stopping a random person on the street and asking, "How many grams of xanthan gum should I put in my suspension?" The chance that they will give you a useful answer is very low—and the same is true for AI. Beginners treat the AI as if it already knows everything about their specific situation, their goals, and their constraints. But AI doesn't read minds. It responds based only on the information you give it. To avoid this mistake, always provide context first. Explain the bigger picture: what you are trying to achieve, what you have already tried, and what the problem really is. Think of the AI as a smart colleague who just joined your team—you wouldn't ask them a highly specific question without first explaining what you're working on. Start with the idea, then ask the question. This small shift in approach makes the difference between a useless answer and a genuinely helpful one.
