Skip to content
Back to all articles
By TJ SoftUpdated:

How to Start Using Artificial Intelligence in 2025 If You Haven’t Used It Before

Jak zacząć korzystać ze sztucznej inteligencji w 2025 roku, jeśli wcześniej tego nie robiłeś

If you are only just discovering tools based on artificial intelligence, you may feel slightly overwhelmed. In 2025, AI is present in phones, computers, work and learning applications, and even household devices. Despite this ubiquity, many people still feel that it is technology that is too complicated to start using consciously. This article will help you enter the world of AI calmly and without feeling pressured, so that from the very first day you can start doing genuinely useful things. I will also show you how to avoid the basic risks. It is important to remember that language models can make mistakes, so content created by AI must be verified; moreover, it is a good idea to take care of your privacy and security. In most cases, this is not too complicated, and it is enough to remember basic rules, such as not sharing private information or passwords with language models that are not hosted locally. You should also be careful with websites offering cheap/free access to a model for other models.

In addition to tips for people who have never used artificial intelligence before, I have also included in the text several interesting pointers that many people who regularly use AI still do not know about. They concern, for example, study mode, Projects mode in ChatGPT, using NotebookLM, creating Excel formulas, and simple programs for personal use. In Grok, you can also generate videos with sound for free; this is also useful when you only need images.

Table of Contents

Introduction: why learn about AI at all, and what are the possibilities of AI.

There are three main reasons to learn about artificial intelligence.

  • AI can significantly speed up your work and save you money.
  • AI may take your job away—so it’s a good idea to know what AI will not be able to do for a long time.
  • If you want to become more involved, AI can help you find a new job creating artificial intelligence.

A Basic Skill: Awareness of What AI Can Easily Be Used For. Just Like the Internet.

The impact of artificial intelligence on society can be compared to the impact the Internet had on the world. It is happening even faster because all the infrastructure needed to use AI, resulting from the spread of the Internet, has already been created.

When it comes to the Internet, the mere habit of searching for advice gave people many possibilities. The awareness that you can solve a problem in 10 minutes and therefore do not need to go to the library and spend half a day searching for information makes life significantly easier.

The same applies to AI. Google helps solve many problems, but it does not always provide useful guidance. Often, where the search engine fails, AI comes to the rescue.

Where to start if you have no experience with AI

The easiest way to start is with tools that require neither installation nor technical knowledge. In 2025, every phone and computer has access to modern language models. You can use them like a search engine, notebook, teacher, or personal assistant. The best-known tools include ChatGPT, Google Gemini, Grok, and Claude.

They differ in their details, but they all work according to a similar principle: you enter a question or instruction, and the system responds in natural language.

If you don't have an app yet, start with the simplest option: the web version. For ChatGPT, simply visit the service's website and create a free account to get started.

How to start using AI without creating accounts (ChatGPT, Grok)

If you don’t like creating new accounts, I have good news for you: you don’t need to log in anywhere to use language models. For ChatGPT and Grok, chat access is available as soon as you visit the website. So all you need to do is search Google for: Grok.com or Chatgpt.com

ChatGPT allows files to be added to a thread without requiring a login, but it does not allow you to choose a specific model, nor can it edit photos.

In Grok’s case, we need to log in to add files, but we can generate images in it without logging in.

The main drawback is that after closing the browser window, all conversations disappear, so it’s a good idea to create an account.

Creating accounts for most models takes literally just a moment. Many of them can be verified using other accounts (e.g. Google, Apple, or Microsoft), so we don’t even have to come up with a password.

If you have Twitter (X), you can immediately use language models (via Grok). If you have a Google account (Gmail), you also have default access to Gemini. LLM chats are also available on WhatsApp and Facebook, although they have limited functionality there.

. .

What to do if you already have an account but have never created anything worthwhile

This is a very common situation. Many people create an account, send two messages, and then close the app because they don’t know what to use it for. The most important thing is to get past this initial stage. You don’t need to create advanced projects right away. To start, it’s best to begin with a casual conversation, through which you can intuitively discover the capabilities of artificial intelligence.

With language models, you can write in Polish.

Some people think that using AI requires writing in English. In fact, there are some tools that require this, but most language models work very well in Polish.

How to Start a Conversation with ChatGPT

You can start a conversation the same way you would start a conversation with a person, for example by asking, “How are you?” However, such responses are generally boring, so it is more interesting, for example, to ask the model whether it is self-aware.

.

We can see that the model is configured to sound like a language model (assistant), rather than an ordinary person, as suggested by the phrase “ready for action.” If we want it to behave more like a human, we need to define the role of the language model. We can do this by writing, for example: act like [a 40-year-old bored with life], or, for instance, “act like Mieszko I.”

.

Such conversations are not just for fun. Defining a role makes it possible to obtain more precise answers in a specific field. It is useful to use the command: “act like a specialist [e.g., a chef, coach, doctor, engineer, mechanic, etc.”

How to Start Using ChatGPT in Practice

Start with small, practical tasks that will genuinely make your life easier.

A few examples you can try right away:

  • ask for an explanation of something you do not understand well enough to explain it to others later;
  • ask for the text you wrote to be corrected;
  • search for information when you do not know the specific words but can describe the problem;
  • send a photo to get a more accurate answer, or when you do not feel like describing something or do not know how to do it;
  • ask for a summary of a longer text (e.g., the post you are currently reading). In addition to the summary itself, you can also ask it questions.

After a few such attempts, you’ll notice that AI works as a universal tool that adapts to your style and needs. You don’t need to know how to program or be familiar with modern technologies.

It is extremely helpful to ask ChatGPT how long a repair might take on your own and whether it can be done yourself or whether it would be better to turn to a specialist right away. You can also describe your skill level.

Just describe the problem, upload a photo (e.g. of a broken door handle, a bicycle part, or shoes coming apart), and add a prompt like: “Can I repair this myself? [I have these skills]—what equipment do I need, and how long will it take?”

This way, you can save time and money while learning new skills. Sometimes, repairing something only requires purchasing an inexpensive, readily available part costing PLN 5, and the repair itself will take 10 minutes. On the other hand, if the repair requires in-depth knowledge and specialized equipment, it is better to ask a professional for help right away rather than struggle for several hours.

Examples of Specific Prompts and Applications:

Help with searching for content, especially if you don’t know what something is called

What is a particular dish called? When I searched on Google, I initially found a dozen results about vermicelli noodles, whereas ChatGPT quickly helped me identify a dish I had eaten once before—“Bun bo nam,” the Vietnamese version I was looking for.

.

We can immediately obtain a recipe for a selected dish. Access to any recipe is nothing new; the internet has offered such possibilities for 30 years. The advantage of ChatGPT is the ability to modify the recipe in such a way that the dish still turns out well.

.

I received a complex recipe for an exquisite dish. Like most of you, I probably don’t feel like making the full version, and I also don’t have oyster sauce. So the question arises: is it essential for this dish? I can also ask ChatGPT to make a version for one person.

.

This can be modified further. Let’s assume that I also don’t have soy sauce. I can therefore ask whether it is essential and what can be used instead. Here, one potential risk becomes apparent: if we ask about substituting every ingredient, we may ultimately end up with a completely different dish, or something that is not good at all. In such cases, it is a good idea to write to the model, for example: “if a given product is essential and cannot be replaced, say so.”

Asking ChatGPT for Photos

If you are preparing food and are not sure whether everything is going according to plan, you can take a photo and ask Chat whether everything is okay. You can use this approach not only when cooking.

. . .

Text Correction and Overcoming Blocks.

Do you know that feeling when you’re writing some text or a message, and you spend 10 minutes on a single passage, only to still have a blank page/empty document in the editor afterward?

Models such as ChatGPT, Grok, Claude, Gemini, and Bielik help overcome this problem. You can simply write a sentence without worrying about errors or grammar and ask the model to rewrite it correctly. You can also ask it to shorten or expand a piece of text. This makes writing much faster. At first, we gain the most when we write most of the text ourselves and ChatGPT helps only when we encounter a block. The proofreading performed by ChatGPT is also very useful.

Texts written entirely by ChatGPT have some applications, but creating them requires more knowledge. In short, in such cases, you can create in 2 hours a text that, without the support of artificial intelligence, would take 12 hours to produce. This way, we still have full control over the content.

Additionally, it is useful to create articles for your own use; in such cases, a single command can produce text that is not of the best quality but describes the exact problem.

If we have a lot of our own content, ChatGPT can create new texts based on existing texts. For example, if we have a lengthy product description, ChatGPT can create a shorter version of it in a few seconds.

.

Films That Clearly Explain the Use of Artificial Intelligence.

Kuba Klawiter, AI from Scratch

This thirty-five-minute video may seem long, but it provides a very good introduction to artificial intelligence. It includes specific, practical examples.

Patomatma How to Use ChatGPT Wisely? (AI – Advice, Experience)

A shorter video, focusing more on how to use ChatGPT for learning.

How to formulate initial prompts so that AI gives meaningful answers

The First Rule of Using ChatGPT – A Casual Conversation

The basic rule is: write as if you were talking to a person. You don’t need to use special commands or artificial language. Guides such as “100 useful prompts” are rarely helpful. Instead, describe what you need clearly and completely. If you want the system to prepare a specific text for you, say who it is intended for, what tone it should have, and what it should include. If you want advice, describe the context as if you were explaining it to a friend.

Sometimes using a specific word or phrase can have a significant impact on the quality of the response, whereas such special prompts are mainly useful when we already have clearly defined needs.

The second principle for using ChatGPT – provide context.

What if a regular conversation with ChatGPT does not produce the expected results? It is important to remember that, unlike an acquaintance or real people, a language model does not know the context. Therefore, it is useful to treat ChatGPT as a specialist from the other side of the world who possesses enormous knowledge but knows about your needs only as much as you tell it.

If you go to a doctor in Poland, even if you say nothing, the mere fact that you visited a clinic conveys some information. Additionally, the doctor can see your condition.

Many people make the mistake of asking one-sentence questions without any background. In such cases, AI may give a very general answer. That’s why it’s worth adding details, even if they seem obvious.

The Third Rule – Not Everything Has to Be in a Single Command

Instead of creating one perfect prompt, it is better to ask the model “on the fly” and then, in the next message, ask it to fix the errors. Later, we can ask the model to create a single prompt based on several messages if we intend to use it more frequently.

Tip – ChatGPT can write prompts for itself.

Asking a language model to write a prompt produces surprisingly good results. Below is a simple text based on a simple prompt:

.

And here you can see the result of the prompt written by ChatGPT.

. .

General Structure of Prompts

If ordinary quick queries do not produce satisfactory answers, it is very useful to use the formula: Role, Topic, Format. For example:

You are a mathematics teacher [Role], Explain to me what integrals are [topic], present it in the form of an article with examples [format].

Each of these three elements can be expanded as needed.

In the case of more complex problems, more parameters can be distinguished. The format can be broken down into style, length, continuous text/list/table, etc.

In specific cases, certain “magic” phrases are useful. For example, “avoid SEO fluff” or “write in terms of benefits” can often produce a better result.

Safety When Using Artificial Intelligence

At the very beginning of using AI, the main risks are language model hallucinations (that is, the risk of AI making errors), as well as data leaks.

How Do AI Hallucinations Arise?

Hallucinations arise for several reasons. One of them is that models do not know the context of a query by default. This was discovered by a sixty-year-old man who asked ChatGPT how to replace salt. The model recommended sodium bromide because, in certain chemical reactions, it can replace sodium chloride, which is table salt.

Conclusion: the information provided by language models needs to be verified more rigorously than information from other sources.

https://www.theguardian.com/technology/2025/aug/12/us-man-bromism-salt-diet-chatgpt-openai-health-information

The second reason for hallucinations stems from how language models are constructed. By default, models do not contain any text or appropriately labeled information. When examining the code of the ChatGPT program, we will not find, for example, information that water boils at a temperature of 100 degrees Celsius. In simple terms, all the knowledge of language models is encoded in the connections between words.

This problem can be solved using your own data. The simplest way is to upload a long text or PDF to the relevant thread. More complex solutions involve creating RAG systems, that is, a kind of knowledge database for language models.

Newer language models have certain safeguards against hallucinations, but they do not work perfectly. AI that uses the internet is precisely one form of RAG. Currently, most larger models can search the internet. However, this is a secondary process; the knowledge from RAG or the internet is not embedded within the language model.

How to Protect Yourself Against Artificial Intelligence Hallucinations?

The primary method is the previously mentioned verification of information provided to language models. This means laborious source checking and verifying correctness.

One shortcut is to paste the answer the model wrote for us and ask in a new thread, “Is this true?”

You can also add an instruction to the prompt such as: “Approach my question critically; if something doesn’t exist or there are no connections, tell me so—don’t make things up.”

Avoid questions that suggest a specific answer

If you are not sure whether something exists, it is a good idea to indicate this in the prompt. The example below demonstrates how the mechanism works. If we ask what happened at a given company on a specific day, the model, instead of saying that it “doesn’t know” or that “most likely nothing noteworthy happened,” will make up some event based on the most common news stories of this type.

Currently, when asked such questions, ChatGPT searches the web and most often provides truthful answers, saying that it did not find anything interesting; however, if we disable internet access, it can still make things up.

.

Particular caution is required in the case of legal inquiries. Here, AI is capable of fabricating regulations.

Generating Images and Videos Using Artificial Intelligence

Currently, creating graphics with AI is much simpler than it was two years ago, when the first widely available image-generation models appeared. Many guides still circulating online describe the process as difficult, even though they are already outdated.

Generating images with AI is a good way to learn how to use artificial intelligence. It is therefore worth experimenting with, even if you do not intend to use it in practice—it is simply more enjoyable than generating text alone, and in this case it is also easier to notice the impact of properly structuring a prompt on the result.

How to Start Generating Images with AI?

All you need to do is enter the command “Create a photo” in ChatGPT, Grok, or Gemini.

.

If for some reason the model generates text instead of an image, you can select the “create image” option at the bottom.

.

The Nano Banana Pro model in Google’s Gemini deserves attention. It is a mix of multiple models; before generating an image, Nano Banana Pro analyzes the content and compares it with factual knowledge. In Google, you can generate ordinary images (and they turn out quite well—Google creates them quickly, too), whereas the main advantage of Google’s new model is precisely its ability to create charts.

A drawback of Gemini is that it sometimes refuses to generate images, especially when the prompt is written in Polish. Additionally, changing the aspect ratio of the generated image is problematic.

.

In this case, Gemini did some conceptual work. Most economies-of-scale charts available online are shaped like the letter “u,” showing the averaged effect of two phenomena.

Modifying AI-Generated Images

AI-generated photos can be easily edited. But not everything can be changed easily. Changing the color of the jacket was trivial; changing the pose produced only a cosmetic effect. A person, upon hearing the remark “a different pose,” would understand that a more substantial change was intended. Of course, this can be worked around by describing the pose we need in detail, or by emphasizing that a major change or, for example, a dynamic pose is intended. But the result is not always satisfactory.

."""

Sometimes a good approach is to use a different photo as an example or paste the image into a new thread.

Uploading photos

Generating images from scratch is fun, but a more useful option is modifying real photos. This makes it possible, for example, to present a product photo in more interesting surroundings. However, it should be remembered that such actions should be illustrative in nature, and where this is not obvious, it is a good idea to emphasize this. It is important not to mislead consumers.

.

In this case, I created 3D visualizations of the logo using Nano Banana.

. .

Films (Grok)

Grok with the Imagine feature enables you to create videos with sound, completely free of charge. Creating a video takes a dozen or so seconds (often faster than creating an image in ChatGPT), the videos have sound, and the free version gives us access to around a dozen videos per day. Videos can be created from scratch based on text, or by pasting in an image.

. .

I believe that the primary use of widely available artificial intelligence is to assist with learning. This can be done in many ways, for example through a regular conversation or ChatGPT’s study mode. In this case, instead of explaining everything, the model asks questions independently and guides the user appropriately.

.

This is often useful, but in my opinion, the Deep Search feature has a much broader range of applications. It is particularly useful when we want to explore a new topic in depth. Traditionally, online research looks like this: you search for a few terms and read the main articles that appear on Google. The first text is generally very useful, the second overlaps with the first by 80%, and the third overlaps with the previous ones by 100%. Most often, the goal is to find a more comprehensive study, but such studies do not exist for every topic, and to find them, you already need to have some basic knowledge.

In practice, reaching the point where you know exactly what you are looking for often takes 30 minutes to 2 hours. At this stage, searching for content is more enjoyable, and having a general understanding makes it easier to absorb.

This is where the deep search feature is particularly helpful. It is available in Grok, Gemini, and ChatGPT; they differ in processing time, complexity, and the length of the results.

."""

In this way, the model searches through a huge number of pages and then extracts the essence from them. As a result, we obtain genuinely useful text in a short time. Creating it often takes a few minutes, sometimes a dozen or several dozen minutes, but in every case, launching it takes a maximum of 2 minutes of our time.

This text is written well enough to be pleasant to read (because we are fairly confident that it is exactly about what we need), although it is generally not suitable for publication.

Reading the result takes 10–20 minutes, and it is much more enjoyable than jumping between random articles. Additionally, you can upload such a text to ElevenLabs and listen to it while doing other things.

Such text is often 80–90% correct. It depends on how well the topic is described and whether we asked the model to search for and gather data or to draw conclusions.

The main advantage of Deep Research is that it provides context. After reading the “research” results, you can easily search Google precisely for specific topics.

.

Dictating Text and Listening to Text

Dictating text is also a major improvement to the work process. Sometimes, when writing text manually, you may encounter writer’s block, and in such a case, articulating your thoughts is often a better solution than writing a “stream of consciousness.” Although writing quickly and then asking AI to create a text based on it is also quite a useful solution.

However, in many cases, dictating text is simply more convenient. And you can do it, for example, while taking a walk. We can dictate the text directly to ChatGPT (and it handles punctuation really well). Sometimes I copy the text directly from the command input field without processing it through the language model.

Another solution is simply to record your voice. You can upload such a recording to Google AI Studio to obtain a transcription. https://turboscribe.ai/

.

The text created in this way then needs to be processed using an LLM, but this is simpler than creating the text from scratch using AI.

Notebook LM and projects

I previously mentioned the possibility of uploading files to language models. If we want to use 10 different texts at once, projects in ChatGPT and NotebookLM are an ideal solution. This allows us to upload multiple texts at once, and every subsequent thread uses the files uploaded initially.

Notebook LM

Notebook LM differs from a typical language model; this tool is used for creating notes and summarizing other content, rather than for conversation itself or creating long texts in a specific style, as is the case with ChatGPT.

."""

Notes Notebook LM consists of three sections: on the left, we have a list of sources (which we previously added ourselves); in the center, there is the chat thread/workspace; and on the right, we have access to additional features (such as a mind map or flashcards), as well as access to saved notes.

**Note!**The main thread in the Notebook is temporary; after refreshing the tab, everything disappears. Therefore, the most important things need to be saved as notes.

The flagship feature of LM Notebook is the ability to freely “turn on and off” sources. This allows us to work with many combinations of texts without the need to laboriously upload and remove them. We can have multiple notebooks, each with different sources.

A particularly useful feature is the ability to add YouTube videos as a source. This allows you to “talk” to videos and summarize them; I use this feature to find useful sections in a video. It only works with videos that have captions enabled, but most newer videos on YouTube have this feature.

.

Summary

In 2025, using AI is as easy and commonplace as searching on Google—it requires no accounts, installations, or knowledge of English. Treat language models like a patient, all-knowing assistant: write in Polish, naturally, provide context, and engage in a casual conversation. From the very first day, you can genuinely save time and money, fix things yourself, find recipes “based on what you have in the fridge,” improve texts, generate photos and videos (for free with sound in Grok), and, thanks to Deep Search, receive a useful analysis of any topic within minutes. Just remember two things: always verify the facts (AI can hallucinate), and don’t upload private data. That’s all. After a few simple attempts, AI will become as natural a tool for you as a smartphone, and it will be hard to imagine life without it.