A hálózat csapdájában

Okay, I admit it, this isn’t a typical IT security topic – but since AI is unavoidable nowadays, it’s better to know about it and not need it rather than the other way around.

The most important things to know before you start using it:

1. First and foremost: AI (whichever ‘model’ you use) is primarily a product, designed to encourage you to use it as much as possible. That is why its most important built-in ‘instruction’ is to ‘serve’ you as well as possible – in other words, to provide what it thinks you want, based on its programmers and the text database it has been fed. This is not necessarily reality!
2. AI does not ‘think’ in the human sense – based on the contents of the gigantic text database it has been fed, it tries as thoroughly as possible to guess what you would like to hear or read based on what you have written to it. It is unable to distinguish between truth and falsehood, because it is incapable of thinking. Based on the text database it has been fed, it tries to respond by matching what you have typed with the ‘best’ of what it has on file, so that it can be used as much as possible.
3. The AI models of the largest companies all work best with their own company’s products – Microsoft Copilot handles Office documents best, Google Gemini handles Google Docs files, and so on.
4. The vast majority of the large text database fed into the AI consists of English-language text or text in related languages – it struggles with longer texts in Hungarian, which is structured according to a completely different internal logic.
   Of course, it will do what you ask, even in Hungarian, but pay close attention to the end result – not just for factual accuracy, but also for linguistic correctness.

What you absolutely need to know before using it:

1. AI is only as smart as its user – remember, its primary purpose is to serve you!
2. You must formulate exactly, precisely, thoroughly and unambiguously what you want it to do. Don’t sacrifice clear comprehensibility on the altar of politeness; the AI won’t be able to read the hidden meaning in your sentences!
3. If you want the AI to do anything for you, your request must absolutely include the following:
   1. Exactly what it should create (e.g. a finished piece, a draft or a brainstorming prompt)
   2. What creative constraints apply (maximum text length, image style, etc.)
   3. What the target audience is (customers, colleagues, company executives, people with only a superficial knowledge of the topic, etc.)
   4. From whose perspective should it be created / In what capacity should it create it (i.e. it should create the work as an expert, as the manager of a small business with an interest in the subject, as a casual internet user who is enthusiastic about the subject, etc.)
   5. What is the purpose of the work (advertising, an overview for other business leaders, an introduction to a series of presentations, an attention-grabbing presentation for those not familiar with the subject, etc.)